aidevops
Advanced tools
AI DevOps Framework - AI-assisted development workflows, code quality, and deployment automation
aidevops.sh is an OpenCode plugin and AI DevOps framework for people who want AI to do useful work across code, infrastructure, business, marketing, content, and creative projects without turning every job into another long, fragile chat.
Most AI tools still leave you doing the hard coordination yourself: finding the right context, choosing a model, protecting secrets, managing branches, watching CI, spotting stuck work, and remembering what went wrong last time. aidevops puts structure around that work so agents can share context, work safely in parallel, spend model budget where it matters, and leave the system better than they found it.
Recommended setup: OpenCode + OpenAI models. GPT-5.5 is the preferred high-capability model for complex agent work; GPT-5.4 mini is the preferred fast, lower-cost model for triage and routine implementation. Claude models (Anthropic) remain fully supported, and other model providers are evaluated from time to time as their quality, latency, and cost profiles change.
"Scope a mission to redesign the landing pages — break it into milestones, dispatch workers in parallel, validate each milestone, and track budget across the whole project."
One conversation, autonomous project delivery, with security, teamwork, token efficiency, and quality control built in.
Founded by Marcus Quinn on 9th November 2025 to help anyone level-up their AI & Open-Source game.
Maximum value for your time and money. aidevops is built for the gap between “the model can probably do this” and “the work is actually done, verified, safe, and worth the cost.”
The result: an AI operations platform that manages projects across every business domain — absorbing everything automatable so you can focus on what matters.
Built on proven patterns: aidevops implements industry-standard agent design patterns - including multi-layer action spaces, context isolation, and iterative execution loops.
Beyond single-task AI. A normal AI harness can be brilliant for one job and still weak at the work around the job. aidevops is for the surrounding discipline: context, routing, safety, git hygiene, collaboration, verification, memory, and follow-through.
What makes it different:
aidevops is MIT licensed. Reuse is welcome, including commercial use, provided the copyright and license notices are retained. If you build a derivative framework, automation bot, workflow library, or commercial product from aidevops code or distinctive operating patterns, see ATTRIBUTION.md for the required notices and preferred credit text.
npm install -g aidevops && aidevops updateaidevops CLI, ~/.aidevops/agents/AGENTS.mdaidevops init - Initialize in any projectaidevops update - Update frameworkaidevops auto-update - Automatic update polling (enable/disable/status)aidevops gpt56-context [enable|disable|status] - Keep GPT-5.6 at a 300K advertised context window in OpenCode (enabled by default), so 80% auto-compaction runs near 240K before long-context pricing; disable to use native provider limitsaidevops secret - Manage secrets (gopass encrypted, AI-safe)aidevops security - Full security assessment (posture, secrets, supply chain)aidevops lint audit - Audit native lint/format/typecheck commands and repo-verify hooksaidevops lint configure --dry-run - Preview safe evidence-based lint provisioningaidevops metrics generate - Generate local LOC, language, and dependency data for README badges or app about pages/onboarding - Interactive setup wizard (in AI assistant)/design-artifact - Route artifact-first UI, deck, email, poster, and mobile mockup work/open-design - Manage the optional Open Design companion studio/auto-browse - Learn, optimize, and graduate repeatable browser operations and web data-mining workflows/report-render - Render report-ready Markdown or JSON to HTML with sticky TOC, print CSS, evidence badges, and source cards for PDF export/report-token-use - Generate a local per-session token, model, compaction, and MCP-use report/pulse - Run the autonomous supervisor loop for dispatch, merge, diagnostics, and stuck-work recovery/serve-sim / serve-sim-helper.sh - Exercise mobile web flows in simulator-backed local previews.agents/scripts//skills recommend "TASK", /onboarding, or the OpenCode agent picker.Since the last README feature refresh, aidevops has added or expanded:
docs/gui/).serve-sim mobile web testing support (.agents/tools/mobile/)..agents/reference/github-self-hosted-runners.md).Comprehensive DevOps framework with tried & tested services integrations, popular and trusted MCP servers, and enterprise-grade infrastructure quality assurance code monitoring and recommendations.
Vault security model: aidevops defines protected data classes, provider
routing labels, trust boundaries, and phased encrypted sync architecture in
.agents/reference/vault.md. Vault guidance is local-first: third-party AI
providers can only reason over data that has been decrypted into their prompt or
tool context, so provider-side logs/retention remain outside Vault's technical
control. Future local LLM mode reduces provider exposure but not local host
compromise risk.
Vault passphrase warning: aidevops cannot recover a lost Vault passphrase. Save it in a trusted password manager with backups; never paste it into AI chat, CLI arguments, environment variables, logs, issue comments, or test fixtures.
Use aidevops to turn evidence bundles into decision-ready reports while keeping Markdown or JSON as the canonical source. Report agents can produce AI-search audits, SEO/GEO scorecards, delivery reviews, campaign reports, board packs, incident summaries, recurring client handoffs, and before/after remediation evidence.
New report capabilities include:
verified, partial, inferred, or missing evidence badges.basic no-CSS output for lightweight handoff.*-a4.pdf, *-usletter.pdf, and *-slides.pdf names._reports/examples/; open _reports/examples/index.html locally to browse the example reports, rendered styles, and PDF exports.Create a report:
reports/general.md for structure, then the matching domain report doc such as reports/seo-geo.md, reports/development.md, reports/marketing.md, or reports/business.md.run: steps or service helpers for collection, then ask the domain agent plus agent:Reports to interpret and prioritise.report.md or report.json in _reports/drafts/<report-name>/ while working, or in _reports/examples/<example-name>/ only after privacy review./report-render report.md or .agents/scripts/report-render-helper.sh render report.md --template <style> --theme auto --pdf-profile a4 --output report.html.Create a repeatable report agent:
reports/routine-handoff.md and tools/build-agent/build-agent.md.run: steps and reserve agent:Reports for narrative, evidence interpretation, recommendations, and handoff tasks.custom/ for local/client-specific agents; shared .agents/ only for broadly reusable framework agents).This framework provides agentic AI assistants with powerful infrastructure access. Use responsibly.
Capabilities: Execute commands, access credentials, modify infrastructure, interact with APIs Your responsibility: Use trusted AI providers, rotate credentials regularly, monitor activity
aidevops security # Run ALL checks (posture + hygiene + supply chain)
aidevops security posture # Interactive security posture setup (gopass, gh auth, SSH)
aidevops security status # Combined posture + hygiene summary
aidevops security scan # Secret hygiene & supply chain scan only
aidevops security scan-pth # Python .pth file audit (supply chain attack vector)
aidevops security scan-secrets # Plaintext credential locations only
aidevops security scan-deps # Unpinned dependency check
aidevops security check # Per-repo security posture assessment
aidevops security dismiss <id> # Dismiss a security advisory after taking action
Running aidevops security with no arguments is the single command that covers everything — user security posture, plaintext secret detection, supply chain IoC scanning, and active advisories.
Security advisories are delivered via aidevops update and shown in the session greeting until dismissed. The scanner never exposes secret values — only file locations and key names. All remediation commands should be run in a separate terminal, not inside AI chat sessions.
Supply chain hardening: All Python dependencies are pinned to exact versions (==) to prevent malicious package upgrades. The .pth file auditor detects known supply chain attack indicators (e.g., the LiteLLM March 2026 PyPI compromise).
npm (recommended - verified provenance):
npm install -g aidevops && aidevops update
Note: npm suppresses postinstall output. The
&& aidevops updatedeploys agents to~/.aidevops/agents/. The CLI will remind you if agents need updating.
Bun (fast alternative):
bun install -g aidevops && aidevops update
Homebrew (macOS/Linux):
brew install marcusquinn/tap/aidevops && aidevops update
Direct from source (aidevops.sh):
bash <(curl -fsSL https://aidevops.sh/install)
Manual (git clone):
git clone https://github.com/marcusquinn/aidevops.git ~/Git/aidevops
~/Git/aidevops/setup.sh
That's it! The setup script will:
~/Git/aidevops~/.aidevops/agents/aidevops CLI commandgit/gh/test/lint command summaries before output reaches AI contextclaude alias that runs claude --dangerously-skip-permissions (skips per-tool permission prompts). Re-running setup updates the alias automatically. To grant permissions per-session instead, press Shift-Tab inside Claude Code to cycle through permission modes (default → skip permissions → auto-approve).New users: Start OpenCode and type /onboarding to configure your services interactively. OpenCode is the recommended tool for aidevops; pair it with OpenAI GPT-5.5 and GPT-5.4 mini for the best current results across agent tiers. The onboarding wizard will:
After installation, use the CLI:
aidevops status # Check what's installed
aidevops doctor # Detect duplicate installs and PATH conflicts
aidevops update # Update framework + check registered projects
aidevops auto-update # Manage automatic update polling (every 10 min)
aidevops init # Initialize aidevops in any project
aidevops features # List available features
aidevops repos # List/add/remove registered projects
aidevops design # DESIGN.md detection + brand guideline HTML/PDF exports
aidevops detect # Scan for unregistered aidevops projects
aidevops upgrade-planning # Upgrade TODO.md/PLANS.md to latest templates
aidevops update-tools # Check and update installed tools
aidevops uninstall # Remove aidevops
aidevops now treats design as a self-contained stack with optional peripherals:
DESIGN.md standard: AI-readable design systems with YAML tokens, linting, previews, and brand/style libraries (.agents/tools/design/design-md.md).aidevops init records has_interface and seeds root DESIGN.md for standard repos plus minimal-scope repos with detected GUI/interface markers.aidevops design guidelines . --pdf generates _reports/brand-guidelines/brand-guidelines.md, HTML, and A4/US Letter/slides PDFs from DESIGN.md.aidevops design survey --json audits owned initialized GUI repos; aidevops design issues --apply files worker-ready auto-dispatch issues for missing DESIGN.md/brand-guideline artifacts./design-artifact decides whether to use aidevops-native implementation or a companion artifact studio; /open-design manages optional Open Design workflows.Optional companion: Open Design by nexu-io (Apache-2.0) is supported as a peripheral for live sandboxed previews, design-skill pickers, .od/ artifact workspaces, and HTML/PDF/PPTX/ZIP-style exports. aidevops remains canonical for agents, skill ingestion, Google DESIGN.md, local hosting, and verification.
# Inspect optional companion status
open-design-helper.sh status
# Print safe install plan only
open-design-helper.sh install
# Install alongside aidevops only after opting in
open-design-helper.sh install --execute
# Start through aidevops local HTTPS if Open Design only prints localhost
open-design-helper.sh start --https-local open-design
# → https://open-design.local when localdev is configured
Imported Open Design skills are not copied verbatim. They are evaluated through aidevops build-agent methodology, deduplicated against existing agents, flattened into aidevops *-skill.md structure, attributed to upstream, and given verification commands. See .agents/tools/design/open-design-ingestion.md for the full skill-value matrix.
Project tracking: When you run aidevops init, the project is automatically registered in ~/.config/aidevops/repos.json. Running aidevops update checks all registered projects for version updates.
Initialize aidevops features in any git repository:
cd ~/your-project
aidevops init # Enable the default feature set
aidevops init planning # Enable only planning
aidevops init planning,time-tracking # Enable specific features
aidevops init deployment-context # Scaffold a deployment manifest
aidevops init wordpress-context # Scaffold WordPress + deployment manifests
This creates:
.aidevops.json - Configuration with enabled features.agents/AGENTS.md - Project-specific agent contextTODO.md - Quick task tracking with time estimatestodo/PLANS.md - Complex execution plans.beads/ - Task graph database (if beads enabled).aidevops/deployments.yaml - Deployment instance inventory (opt-in).aidevops/wordpress.yaml - WordPress and LocalWP context (opt-in)Available features: planning, git-workflow, code-quality, time-tracking, database, beads, sops, security, deployment-context, wordpress-context. hosting-context is an alias for deployment-context; wordpress-context implies deployment context. The default all set intentionally excludes both project-instance context features.
After aidevops init registers a new repo, run /setup-git in your AI assistant
to apply per-repo platform secrets. Most notably, this sets SYNC_PAT — a
GitHub Actions secret that lets issue-sync.yml push TODO.md auto-completion
past branch protection.
This is distinct from /onboarding (per-account credentials like gh auth login):
GitHub Actions secrets are scoped per-repo, so each repo needs its own. You need
gh auth login to succeed before any per-repo helper can run, so /onboarding
comes first, /setup-git second.
Run /setup-git again whenever you register a new repo with aidevops repos add
or when a SYNC_PAT advisory appears in the session greeting toast. If you skip
this step, issue-sync.yml will post a remediation comment when it hits branch
protection — /setup-git walks through the fix.
When aidevops templates evolve, upgrade existing projects to the latest format:
aidevops upgrade-planning # Interactive upgrade with backup
aidevops upgrade-planning --dry-run # Preview changes without modifying
aidevops upgrade-planning --force # Skip confirmation prompt
This preserves your existing tasks while adding TOON-enhanced parsing, dependency tracking, and better structure.
Automatic detection: aidevops update now scans all registered projects for outdated planning templates (comparing TOON meta version numbers) and offers to upgrade them in-place with backups.
Beads provides task dependency tracking and graph visualization:
aidevops init beads # Enable beads (includes planning)
Task Dependencies:
- [ ] t001 First task
- [ ] t002 Second task blocked-by:t001
- [ ] t001.1 Subtask of t001
| Syntax | Meaning |
|---|---|
blocked-by:t001 | Task waits for t001 to complete |
blocks:t002 | This task blocks t002 |
t001.1 | Subtask of t001 (hierarchical) |
Commands:
| Command | Purpose |
|---|---|
/ready | Show tasks with no open blockers |
/list-verify | List verification queue (pending, passed, failed) |
/sync-beads | Sync TODO.md/PLANS.md with Beads graph |
bd list | List all tasks in Beads |
bd ready | Show ready tasks (Beads CLI) |
bd graph <id> | Show dependency graph for an issue |
Architecture: aidevops markdown files (TODO.md, PLANS.md) are the source of truth. Beads syncs from them for visualization.
Optional Viewers: Beyond the bd CLI, there are community viewers for richer visualization:
beads_viewer (Python TUI) - PageRank, critical path analysisbeads-ui (Web) - Live updates in browserbdui (React/Ink TUI) - Modern terminal UIperles (Rust TUI) - BQL query languageSee .agents/tools/task-management/beads.md for complete documentation and installation commands.
Your AI assistant now has agentic access to 30+ service integrations.
OpenCode with OpenAI is the current recommended aidevops setup. Use GPT-5.5 for complex reasoning, architecture, security-sensitive review, and hard agent tiers; use GPT-5.4 mini for fast triage, routine implementation, retries, and lower-cost worker throughput.
Authenticate via the pool:
aidevops model-accounts-pool add openai
# Restart OpenCode after adding
Why this is the default:
OpenCode includes Anthropic OAuth authentication natively — no API key needed. OAuth is covered by your Claude Pro/Max subscription at zero additional cost.
Authenticate via the pool (recommended):
aidevops model-accounts-pool add anthropic
# Opens browser OAuth flow — no API key required
# Restart OpenCode after adding
Or via the OpenCode TUI:
Open OpenCode → Ctrl+A → Select Anthropic → Login with Claude.ai → follow browser OAuth flow.
Note:
opencode auth loginprompts for an API key, not OAuth. Use the commands above for subscription-based OAuth access.
Benefits:
Access Cursor Pro models (Composer 2, Claude 4.6 Opus/Sonnet, GPT-5.x, Gemini 3.1 Pro) in OpenCode through a local gRPC proxy that translates OpenAI-compatible requests to Cursor's protobuf/HTTP2 protocol.
Setup:
# Add your Cursor account to the pool (reads from local Cursor IDE)
oauth-pool-helper.sh add cursor
# Restart OpenCode — Cursor models appear in Ctrl+T model picker
How it works:
Benefits:
Use your Google AI Pro, AI Ultra, or Workspace subscription for Gemini models. Tokens are injected as ADC bearer tokens that Gemini CLI, Vertex AI SDK, and the Gemini API pick up automatically.
Setup:
# Add your Google account to the pool (browser OAuth flow)
aidevops model-accounts-pool add google
# Restart OpenCode — token is injected as GOOGLE_OAUTH_ACCESS_TOKEN
Supported plans:
Isolation guarantee: Google auth failures never affect Anthropic/OpenAI/Cursor providers. A Google 429 or auth error only puts the Google pool into cooldown.
Enable AI-powered issue resolution directly from GitHub. Comment /oc fix this on any issue and the AI creates a branch, implements the fix, and opens a PR.
Security-first design - The workflow includes:
ai-approved label required on issues before AI processingQuick setup:
# 1. Install the OpenCode GitHub App
# Visit: https://github.com/apps/opencode-agent
# 2. Add API key secret for your chosen provider
# Repository → Settings → Secrets → OPENAI_API_KEY or ANTHROPIC_API_KEY
# 3. Create required labels
gh label create "ai-approved" --color "0E8A16" --description "Issue approved for AI agent"
gh label create "security-review" --color "D93F0B" --description "Requires security review"
The secure workflow is included at .github/workflows/opencode-agent.yml.
Usage:
| Context | Command | Result |
|---|---|---|
Issue (with ai-approved label) | /oc fix this | Creates branch + PR |
| Issue | /oc explain this | AI analyzes and replies |
| PR | /oc review this PR | Code review feedback |
| PR Files tab | /oc add error handling here | Line-specific fix |
See .agents/tools/git/opencode-github-security.md for the full security documentation.
Supported AI tool: OpenCode is the recommended and tested AI coding tool for aidevops. All features, agents, and workflows are designed and tested for OpenCode first. We recommend OpenAI models for the best current results across all agent tiers: GPT-5.4 mini for fast triage/routine work and GPT-5.5 for complex implementation, review, and reasoning. Claude models (Anthropic) remain fully supported, and other providers are tested as their capabilities change.
Recommended stack:
If you see "Anthropic Key Missing", "OpenAI Key Missing", or the model stops responding, run these commands from any terminal — no working model session required.
Step 1 — Check pool health
aidevops model-accounts-pool status # counts: available / rate-limited / auth-error
aidevops model-accounts-pool check # live token validity test per account
Step 2 — Fix based on what you see
| Symptom | Command |
|---|---|
OpenAI account shows rate-limited | aidevops model-accounts-pool rotate openai |
Anthropic account shows rate-limited | aidevops model-accounts-pool rotate anthropic |
| All accounts in cooldown | aidevops model-accounts-pool reset-cooldowns |
OpenAI account shows auth-error | aidevops model-accounts-pool add openai (re-auth) |
Anthropic account shows auth-error | aidevops model-accounts-pool add anthropic (re-auth) |
| Pool is empty (no accounts) | aidevops model-accounts-pool add openai |
| Recently re-authenticated, still broken | aidevops model-accounts-pool assign-pending openai |
| Google Gemini CLI rate-limited | aidevops model-accounts-pool rotate google |
| Google token expired | aidevops model-accounts-pool add google (re-auth) |
Step 3 — If still broken, re-add the account
aidevops model-accounts-pool add openai # ChatGPT Plus/Pro
aidevops model-accounts-pool add anthropic # Claude Pro/Max — opens browser OAuth
aidevops model-accounts-pool add cursor # Cursor Pro (reads from local IDE)
aidevops model-accounts-pool add google # Google AI Pro/Ultra/Workspace — browser OAuth
aidevops model-accounts-pool import claude-cli # Import from existing Claude CLI auth
Restart OpenCode after any add, rotate, or reset-cooldowns to pick up the new credentials.
Full command reference
aidevops model-accounts-pool status # Pool health at a glance
aidevops model-accounts-pool list # Per-account detail + expiry
aidevops model-accounts-pool check # Live API validity test
aidevops model-accounts-pool rotate [provider] # Switch to next available account NOW
aidevops model-accounts-pool reset-cooldowns # Clear all rate-limit cooldowns
aidevops model-accounts-pool assign-pending <p># Assign stranded pending token
aidevops model-accounts-pool remove <p> <email># Remove an account
Note:
reset-cooldownsclears cooldowns in the pool file. If OpenCode is already running, the in-memory token endpoint cooldown is only cleared when OpenCode restarts or when you use the/model-accounts-pool reset-cooldownsslash command inside an active session.
If you prefer guided help: Open OpenCode with a free model (OpenCode Zen includes free models that don't require any API key or subscription) and run the auth troubleshooting agent by typing:
@auth-troubleshooting
The agent contains the full recovery flow and symptom table. Free models work fine for this — no paid subscription needed.
Your terminal tab/window title automatically shows repo/branch context when working in git repositories. Interactive OpenCode tabs add ⚪ as soon as the first root-session message is submitted and keep it when the initial title becomes descriptive, 🔴 while retrying after an error, 🟡 while awaiting permission, and 🟢 when finished and awaiting input. This helps identify both the work context and session state across multiple terminal sessions.
Supported terminals: Tabby, cmux, iTerm2, Kitty, Alacritty, WezTerm, Hyper, and most xterm-compatible terminals.
How it works: The pre-edit-check.sh script's primary role is enforcing git workflow protection (blocking edits on main/master branches). As a secondary, non-blocking action, it updates the terminal title via escape sequences. The OpenCode plugin listens for root-session user-message, status, and permission events and updates the same title without changing the stored OpenCode session name. No configuration is needed when dynamic terminal titles are enabled.
Example format: {repo}/{branch-type}/{description}
See .agents/tools/terminal/terminal-title.md for customization options.
Companion tool:
Collaborator compatibility: Projects initialized with aidevops init include pointer files (.cursorrules, .windsurfrules, etc.) that reference AGENTS.md, helping collaborators using other editors find project context. aidevops does not install into or configure those tools.
Repo courtesy files: aidevops init scaffolds standard repo files if they don't exist: DESIGN.md for GUI/interface repos, README.md, LICENCE (MIT), CHANGELOG.md, CONTRIBUTING.md, SECURITY.md, CODE_OF_CONDUCT.md. Author name and email are auto-detected from git config. Existing files are never overwritten.
AI-First Infrastructure Management:
Autonomous Orchestration:
/mission scopes a high-level goal into milestones and features. The pulse dispatches workers, validates milestones, tracks budget, and advances through the project automatically (mission-dashboard-helper.sh)/runners command - Batch dispatch from task IDs, PR URLs, or descriptions with concurrency control and progress monitoringbudget-tracker-helper.sh) with burn-rate analysis and /budget-analysis command for model routing decisionsProject Intelligence:
bundle-helper.sh)ttsr-rule-loader.sh) with .agents/rules/ directory for AI output correction (e.g., no-edit-on-main, no-glob-for-discovery)/cross-review dispatches the same prompt to multiple AI models in parallel, diffs results, and optionally auto-scores via a judge modellocal-model-helper.sh) with HuggingFace GGUF model management/tech-stack detects technology stacks of URLs or finds sites using specific technologies (Wappalyzer, httpx, nuclei, BuiltWith)ip-reputation-helper.sh checks IP addresses against multiple reputation databases (Spamhaus, ProxyCheck, AbuseIPDB) before VPS purchase or deploymentserve-simConversational Memory & Entity System:
entity-helper.sh): people, agents, and services tracked across Matrix, SimpleX, email, and CLI with versioned profilesconversation-helper.sh): idle detection, immutable summaries, tone profile extractionCommunications:
simplex-bot/)matterbridge-helper.sh)xurl workflows for search, timelines, bookmarks, posting, replies, DMs, media, and raw API reads. Supports multiple X developer apps/subscription tiers with --app and multiple authenticated accounts with --username; model-provider auth such as OpenCode xAI/Grok remains separate from X API OAuth (content/social-xurl.md, xurl-helper.sh).local domains with HTTPS (localdev-helper.sh)MCP Toolkit:
mcporter npm package)Unified Interface:
Quality Control & Monitoring:
aidevops review-gate owner/repo wait) or allow merge with follow-up quality coverage (aidevops review-gate owner/repo pass, the default). Per-tool overrides are supported, for example aidevops review-gate owner/repo --tool coderabbitai wait. Failed, skipped, or placeholder bot states are not treated as rate limits and continue to block until a real review/status appears or a human resolves them./performance command)aidevops includes curated skills imported from external sources. Skills support automatic update tracking:
| Skill | Source | Description |
|---|---|---|
| cloudflare-platform | dmmulroy/cloudflare-skill | 60 Cloudflare products: Workers, Pages, D1, R2, KV, Durable Objects, AI, networking, security |
| heygen | heygen-com/skills | AI avatar video creation API: avatars, voices, video generation, streaming, webhooks |
| remotion | remotion-dev/skills | Programmatic video creation with React, animations, rendering |
| video-prompt-design | snubroot/Veo-3-Meta-Framework | AI video prompt engineering - 7-component meta prompt framework for Veo 3 |
| animejs | animejs.com | JavaScript animation library patterns and API (via Context7) |
| caldav-calendar | ClawdHub | CalDAV calendar sync via vdirsyncer + khal (iCloud, Google, Fastmail, Nextcloud) |
| proxmox-full | ClawdHub | Complete Proxmox VE hypervisor management via REST API |
CLI Commands:
aidevops skill add <owner/repo> # Import a skill from GitHub
aidevops skill add clawdhub:<slug> # Import a skill from ClawdHub
aidevops skill list # List imported skills
aidevops skill check # Check for upstream updates
aidevops skill update [name] # Update specific or all skills
aidevops skill scan [name] # Security scan skills (Cisco Skill Scanner)
aidevops skill remove <name> # Remove an imported skill
Skills are registered in ~/.aidevops/agents/configs/skill-sources.json with upstream tracking for update detection.
Security Scanning:
Imported skills are automatically security-scanned using Cisco Skill Scanner when installed. Scanning runs on both initial import and updates -- pulling a new version of a skill triggers the same security checks as the first import. CRITICAL/HIGH findings block the operation; MEDIUM/LOW findings warn but allow. Telemetry is disabled - no data is sent to third parties.
When a VirusTotal API key is configured (aidevops secret set VIRUSTOTAL_MARCUSQUINN), an advisory second layer scans file hashes against 70+ AV engines and checks domains/URLs referenced in skill content. VT scans are non-blocking -- the Cisco scanner remains the security gate.
| Scenario | Security scan runs? | CRITICAL/HIGH blocks? |
|---|---|---|
aidevops skill add <source> | Yes | Yes |
aidevops skill update [name] | Yes | Yes |
aidevops skill add <source> --force | Yes | Yes |
aidevops skill add <source> --skip-security | Yes (reports only) | No (warns) |
aidevops skill scan [name] | Yes (standalone) | Report only |
The --force flag only controls file overwrite behavior (replacing an existing skill without prompting). To bypass security blocking, use --skip-security explicitly -- this separation ensures that routine updates and re-imports never silently skip security checks.
Scan results are logged to .agents/SKILL-SCAN-RESULTS.md automatically on each batch scan and skill import, providing a transparent audit trail of security posture over time.
Browse community skills: skills.sh | ClawdHub | Specification: agentskills.io
Reference:
Sync agents from private Git repositories into the framework. Private repos keep their own agents, helper scripts, and slash commands — aidevops sources sync deploys them alongside the core agents.
aidevops sources add ~/Git/my-private-agents # Register a local repo
aidevops sources add-remote git@github.com:u/r.git # Clone and register a remote repo
aidevops sources list # List configured sources
aidevops sources sync # Sync all sources
aidevops sources remove my-private-agents # Remove a source
How it works: Private repos contain a .agents/ directory with agent subdirectories. Agents with mode: primary in their frontmatter are symlinked to the agents root for auto-discovery as primary agent tabs. Markdown files with agent: frontmatter are deployed as /slash commands. All sources sync automatically during aidevops update.
Reference: .agents/aidevops/agent-sources.md
aidevops implements proven agent design patterns identified by Lance Martin (LangChain).
| Pattern | Description | aidevops Implementation |
|---|---|---|
| Give Agents a Computer | Filesystem + shell for persistent context | ~/.aidevops/.agent-workspace/, 1,630+ helper scripts |
| Multi-Layer Action Space | Few tools, push actions to computer | Per-agent MCP filtering (~12-20 tools each) |
| Knowledge Graph Routing | Indexed, cross-referenced agents instead of isolated skills | subagent-index.toon maps 2,150+ agents by domain, purpose, and dependency — agents discover related context through the graph, not just their own file |
| Progressive Disclosure | Load context on-demand | Subagent routing with content summaries, YAML frontmatter, read-on-demand |
| Offload Context | Write results to filesystem | .agent-workspace/work/[project]/ for persistence |
| Cache Context | Prompt caching for cost | Stable instruction prefixes |
| Isolate Context | Sub-agents with separate windows | Subagent files with specific tool permissions |
| Multi-Agent Orchestration | Coordinate parallel agents | TOON mailbox, agent registry, supervisor dispatch |
| Compaction Resilience | Preserve context across compaction | OpenCode plugin injects dynamic state at compaction time |
| Ralph Loop | Iterative execution until complete | /full-loop, full-loop-helper.sh |
| Evolve Context | Learn from sessions | /remember, /recall with SQLite FTS5 + opt-in semantic search |
| Pattern Tracking | Learn what works/fails | /patterns command, memory-helper.sh |
| Token-Efficient Serialisation | Minimise context overhead for structured data | TOON format — 20-60% token reduction vs JSON/YAML for agent indexes, registries, and data exchange |
| Token-Efficient Tool Output | Summarise noisy terminal output without hiding evidence | RTK is installed by default during setup; start with rtk-helper.sh for compact supported summaries, then rerun raw/direct commands when filtered output is insufficient; bypass compression for file reads, JSON assertions, exact diffs, security scans, and other verbatim evidence |
| Cost-Aware Routing | Match model to task complexity | model-routing.md with provider-aware tier guidance, /route command |
| Model Comparison | Compare models side-by-side | /compare-models (live data), /compare-models-free (offline) |
| Response Scoring | Evaluate actual model outputs | /score-responses with structured criteria |
Key insight: Context is a finite resource with diminishing returns. aidevops treats every token as precious - loading only what's needed, when it's needed.
See .agents/aidevops/architecture.md for detailed implementation notes and references.
Run multiple AI agents in parallel on separate branches, coordinated through a lightweight mailbox system. Each agent works independently in its own git worktree while the supervisor manages task distribution and status reporting.
Architecture:
Supervisor (pulse loop)
├── Agent Registry (TOON format - who's active, what branch, idle/busy)
├── Mailbox System (SQLite WAL-mode, indexed queries)
│ ├── task_assignment → worker inbox
│ ├── status_report → coordinator outbox
│ └── broadcast → all agents
└── Model Routing (tier-based: GPT-5.4 mini / GPT-5.5 / provider fallbacks)
Key components:
| Component | Script | Purpose |
|---|---|---|
| Mailbox | mail-helper.sh | SQLite-backed inter-agent messaging (send, check, broadcast, archive) |
| Supervisor | supervisor-helper.sh | Autonomous multi-task orchestration with SQLite state machine, batches, retry cycles, cron scheduling, auto-pickup from TODO.md |
| Registry | mail-helper.sh register | Agent registration with role, branch, worktree, heartbeat |
| Model routing | model-routing.md, /route | Cost-aware routing across OpenAI, Anthropic, Gemini, Cursor, Grok, and local providers |
| Budget tracking | budget-tracker-helper.sh | Append-only cost log for model routing decisions |
| Observability | observability.mjs, runtime-events.mjs | LLM/tool metrics plus redacted lifecycle, lineage, and state evidence in one local SQLite database |
How it works:
mail-helper.sh register --role worker)supervisor-helper.sh pulse)busy_timeout handles concurrent access (79x faster than previous file-based system)Compaction plugin (.agents/plugins/opencode-aidevops/): When OpenCode compacts context (at ~200K tokens), the plugin injects current session state - agent registry, pending mailbox messages, git context, and relevant memories - ensuring continuity across compaction boundaries.
Custom system prompt (.agents/prompts/build.txt): Based on upstream OpenCode with aidevops-specific overrides for tool preferences, professional objectivity, and per-model reinforcements for weaker models.
Subagent index (.agents/subagent-index.toon): Compressed TOON routing table listing all agents, subagents, workflows, and scripts with model tier assignments - enables fast agent discovery without loading full markdown files.
Why this matters: Long-running tasks -- batch PR reviews, multi-site audits, large refactors, multi-day feature projects -- are where AI agents deliver the most value. Instead of babysitting one task at a time, the supervisor dispatches work to parallel agents, each in its own git worktree, with automatic retry, progress tracking, and batch completion reporting.
The pulse is the heartbeat of aidevops — an autonomous AI supervisor that runs every 2 minutes via launchd. There is no human at the terminal. It manages the entire development pipeline across all repos registered with pulse: true.
What it does each cycle:
| Phase | Action |
|---|---|
| Capacity check | Circuit breaker, dynamic worker slots calculated from available RAM |
| Merge ready PRs | Green CI + no blocking reviews → squash merge (free — no worker slot needed) |
| Fix failing PRs | Dispatch a worker to fix CI failures or address review feedback |
| Detect stuck work | PRs open 6+ hours with no activity → flag or close and re-file |
| Dispatch workers | Route open issues to available worker slots, respecting priority and blocked-by: dependencies |
| Advance missions | Check active multi-day missions, dispatch features, validate milestones, track budget |
| Triage quality | Read daily quality sweep findings (ShellCheck, SonarCloud, Codacy, CodeRabbit), create issues for actionable findings |
| Sync TODOs | Create GitHub issues for unsynced TODO entries, commit ref changes |
| Respect API budget | Use cached/prefetched GitHub metadata, cooldown headers, and ramp pacing before spending more API calls |
| Kill stuck workers | Workers running 3+ hours with no PR are killed to free slots |
| Detect orphaned PRs | Open PRs with no active worker and no activity for 6+ hours are flagged for re-dispatch |
Operational intelligence:
messages / max(1, commits) for each active worker. High ratio (>30) with >30 min elapsed and zero commits flags the worker as "struggling". Ratio >50 after 1 hour flags "thrashing". Informational signal — the supervisor LLM decides the action (kill, wait, re-dispatch with more context).agents/reference/repository-campaigns.mdThe pulse is an LLM, not a script. It reads issue bodies, assesses context, and uses judgment. When it encounters something unexpected — an issue body that says "completed", a task with no clear description, a label that doesn't match reality — it handles it the way a competent human manager would.
# Pulse runs automatically via launchd (every 2 minutes)
# Manual trigger:
opencode run "/pulse"
See: .agents/scripts/commands/pulse.md for the full supervisor specification.
Missions are the highest-level orchestration primitive — autonomous multi-day projects that break a high-level goal into milestones, features, and validation criteria. The pulse supervisor advances them automatically.
# Scope a mission interactively
/mission "Redesign the landing pages for mobile-first with A/B testing"
How missions work:
/mission scopes the goal into milestones with features and acceptance criteriamission:mNNN with a GitHub issueMAX_WORKERS)Two execution modes:
| Mode | Workflow | Best for |
|---|---|---|
| Full | Worktree + PR per feature, standard review flow | Production code, collaborative projects |
| POC | Direct commits, skip ceremony | Prototypes, experiments, proof-of-concept |
Mission state is tracked in a JSON file committed to the repo. Each pulse cycle reads the state, acts on it, and commits updates — so any session (or the next pulse) can pick up where the last one left off.
See: .agents/workflows/mission-orchestrator.md for the full orchestrator specification, .agents/scripts/commands/dashboard.md for the mission progress dashboard.
High-stakes operations are verified by a second AI model from a different provider before execution. This catches single-model hallucinations before destructive operations cause irreversible damage.
When verification triggers:
| Risk Level | Examples | Action |
|---|---|---|
| Critical | git push --force to main, DROP DATABASE, production deploy | Blocked unless second model agrees |
| High | Force push to task ref, data migration, secret exposure | Warned, verification recommended |
| Medium | Bulk file deletion, config changes | Logged |
| Low | Normal edits, test runs | No verification |
How it works:
pre-edit-check.sh screens operations against the high-stakes taxonomyverify-operation-helper.sh sends the operation context to a second model (different provider than the primary)Why cross-provider? Same-provider models share training data and failure modes. A GPT hallucination is unlikely to be reproduced by Claude or Gemini, and vice versa. The verifier uses the cheapest suitable model tier, so cost is minimal per check.
Configuration: Per-repo via .agents/reference/high-stakes-operations.md. Opt-out with VERIFY_ENABLED=false (not recommended).
See: .agents/tools/verification/parallel-verify.md for the verification agent specification.
Bundles are project-type presets that auto-configure model tiers, quality gates, and agent routing per repo. Instead of manually configuring each project, bundles detect what kind of project you're working on and apply sensible defaults.
Built-in bundles:
| Bundle | Auto-detected by | Model default | Quality gates | Agent routing |
|---|---|---|---|---|
web-app | package.json + framework markers | standard | Full (lint, test, build, a11y) | Build+ default |
library | package.json with main/exports | standard | Full + API docs check | Build+ default |
cli-tool | bin field in package.json | standard | ShellCheck, test | Build+ default |
content-site | CMS markers, wp-config.php | fast | Lighthouse, SEO | Marketing for content tasks |
infrastructure | Dockerfile, terraform/, ansible/ | standard | ShellCheck, security scan | Build+ default |
agent | AGENTS.md, .agents/ | thinking | Agent review, prompt quality | Build+ default |
Resolution priority: Explicit bundle field in repos.json > .aidevops.json project config > auto-detection from marker files.
CLI:
bundle-helper.sh detect <repo-path> # Auto-detect bundle type
bundle-helper.sh resolve <repo-path> # Show resolved config (with overrides)
bundle-helper.sh show <bundle-name> # Show bundle defaults
bundle-helper.sh list # List all available bundles
See: .agents/bundles/ for bundle definitions, .agents/scripts/bundle-helper.sh for the CLI.
Run multiple AI sessions concurrently with isolated contexts. Named runners provide persistent agent identities with their own instructions and memory.
| Feature | Description |
|---|---|
| Headless dispatch | headless-runtime-helper.sh run provides canonical model routing, retries, and session isolation |
| Durable one-shot scheduling | aidevops schedule once queues private, restart-safe delayed work without fake recurring jobs or sleeper processes |
| Runners | Named agent instances with per-runner AGENTS.md, config, and run logs (runner-helper.sh) |
| Self-hosted runner runbooks | GitHub runner storage, lifecycle, Docker foreground mode, timer freshness, and cleanup-race guidance (.agents/reference/github-self-hosted-runners.md) |
| Session management | Resume sessions with -s <id> or -c, fork with SDK |
| Memory namespaces | Per-runner memory isolation with shared access when needed |
| SDK orchestration | @opencode-ai/sdk for TypeScript parallel dispatch via Promise.all |
| Matrix integration | Chat-triggered dispatch via self-hosted Matrix (optional) |
# Create a named runner
runner-helper.sh create code-reviewer --description "Reviews code for security and quality"
# Dispatch a task (one-shot)
runner-helper.sh run code-reviewer "Review src/auth/ for vulnerabilities"
# Queue one delayed execution and inspect its lifecycle
aidevops schedule once --after 2h --name "Review authentication" \
--dir ~/Git/example --prompt-file ~/.config/aidevops/prompts/auth-review.md
aidevops schedule status
# Dispatch against warm server (faster, no MCP cold boot)
opencode serve --port 4096 &
runner-helper.sh run code-reviewer "Review src/auth/" --attach http://localhost:4096
# Parallel dispatch via CLI
opencode run --attach http://localhost:4096 --title "Review" "Review src/auth/" &
opencode run --attach http://localhost:4096 --title "Tests" "Generate tests for src/utils/" &
wait
# List runners and status
runner-helper.sh list
runner-helper.sh status code-reviewer
Architecture:
OpenCode Server (opencode serve)
├── Session 1 (runner/code-reviewer)
├── Session 2 (runner/seo-analyst)
└── Session 3 (scheduled-task)
↑
HTTP API / SSE Events
↑
┌────────┴────────┐
│ Dispatch Layer │ ← runner-helper.sh, cron, Matrix bot, SDK
└─────────────────┘
Example runner templates: code-reviewer, seo-analyst - copy and customize for your own runners.
Matrix bot dispatch (optional): Bridge Matrix chat rooms to runners for chat-triggered AI. Each room maintains persistent conversation context via SQLite -- on idle timeout, the session is compacted (summarised) and stored, so the next message resumes with full context.
# Setup Matrix bot (interactive wizard)
matrix-dispatch-helper.sh setup
# Map rooms to runners (each room = separate session)
matrix-dispatch-helper.sh map '!dev-room:server' code-reviewer
matrix-dispatch-helper.sh map '!seo-room:server' seo-analyst
# Start bot (daemon mode)
matrix-dispatch-helper.sh start --daemon
# In Matrix room: "!ai Review src/auth.ts for security issues"
# Manage sessions
matrix-dispatch-helper.sh sessions list
matrix-dispatch-helper.sh sessions stats
See: headless-dispatch.md for full documentation including parallel vs sequential decision guide, SDK examples, CI/CD integration, and custom agent configuration. matrix-bot.md for Matrix bot setup including Cloudron Synapse guide and session persistence.
Agents that learn from experience and contribute improvements:
| Phase | Description |
|---|---|
| Review | Analyze memory for success/failure patterns (memory-helper.sh) |
| Refine | Generate and apply improvements to agents |
| Test | Validate in isolated OpenCode sessions |
| PR | Contribute to community with privacy filtering |
Safety guardrails:
Test agent behavior through isolated AI sessions with automated validation:
# Create a test suite
agent-test-helper.sh create my-tests
# Run tests (auto-detects claude or opencode CLI)
agent-test-helper.sh run my-tests
# Quick single-prompt test
agent-test-helper.sh run-one "What tools do you have?" --expect "bash"
# Before/after comparison for agent changes
agent-test-helper.sh baseline my-tests # Save current behavior
# ... modify agents ...
agent-test-helper.sh compare my-tests # Detect regressions
Test suites are JSON files with prompts and validation rules (expect_contains, expect_not_contains, expect_regex, min_length, max_length). Results are saved for historical tracking.
See: agent-testing.md subagent for full documentation and example test suites.
Speak naturally to your AI coding agent and hear it respond. The voice bridge connects your microphone to OpenCode via a fast local pipeline -- ask questions, give instructions, execute tasks, all by voice.
Mic → Silero VAD → Whisper MLX (1.4s) → OpenCode (4-6s) → Edge TTS (0.4s) → Speaker
Round-trip: ~6-8 seconds on Apple Silicon. The agent can edit files, run commands, create PRs, and confirm what it did -- all via voice.
Quick start:
# Start a voice conversation (installs deps automatically)
voice-helper.sh talk
# Choose engines and voice
voice-helper.sh talk whisper-mlx edge-tts en-GB-SoniaNeural
voice-helper.sh talk whisper-mlx macos-say # Offline mode
# Utilities
voice-helper.sh devices # List audio input/output devices
voice-helper.sh voices # List available TTS voices
voice-helper.sh benchmark # Test STT/TTS/LLM speeds
voice-helper.sh status # Check component availability
Features:
| Feature | Details |
|---|---|
| Swappable STT | whisper-mlx (fastest on Apple Silicon), faster-whisper (CPU) |
| Swappable TTS | edge-tts (best quality), macos-say (offline), facebookMMS (local) |
| Voice exit | Say "that's all", "goodbye", "all for now" to end naturally |
| STT correction | LLM sanity-checks transcription errors before acting (e.g. "test.txte" → "test.txt") |
| Task execution | Full tool access -- edit files, git operations, run commands |
| Session handback | Conversation transcript output on exit for calling agent context |
| TUI compatible | Graceful degradation when launched from AI tool's Bash (no tty) |
How it works: The bridge uses opencode run --attach to connect to a running OpenCode server for low-latency responses (~4-6s vs ~30s cold start). It automatically starts opencode serve if not already running.
Requirements: Apple Silicon Mac (for whisper-mlx), Python 3.10+, internet (for edge-tts). The voice helper installs Python dependencies automatically into the S2S venv.
For advanced use cases (custom LLMs, server/client deployment, multi-language, phone integration), the full huggingface/speech-to-speech pipeline is also available:
speech-to-speech-helper.sh setup # Install pipeline
speech-to-speech-helper.sh start --local-mac # Run on Apple Silicon
speech-to-speech-helper.sh start --cuda # Run on NVIDIA GPU
speech-to-speech-helper.sh start --server # Server mode (remote clients)
Supported languages: English, French, Spanish, Chinese, Japanese, Korean (auto-detect or fixed).
Additional voice methods:
| Method | Description |
|---|---|
| VoiceInk + Shortcut | macOS: transcription → OpenCode API → response |
| iPhone Shortcut | iOS: dictate → HTTP → speak response |
| Pipecat STS | Full voice pipeline: Soniox STT → AI → Cartesia TTS |
See: speech-to-speech.md for full component options, CLI parameters, and integration patterns (Twilio phone, video narration, voice-driven DevOps).
Cron-based agent dispatch for automated workflows:
# Example: Daily SEO report at 9am
0 9 * * * ~/.aidevops/agents/scripts/runner-helper.sh run "seo-analyst" "Generate daily SEO report"
See: TODO.md tasks t109-t118 for implementation status.
aidevops itself is lightweight (shell scripts + markdown), but AI model workloads benefit from capable hardware:
| Tier | Machine | CPU | RAM | GPU | Best For |
|---|---|---|---|---|---|
| Minimum | Any modern laptop | 4+ cores | 8GB | None | Framework only, cloud AI APIs |
| Recommended | Mac Studio / desktop | Apple M1+ or 8+ cores | 16GB+ | MPS (Apple) or NVIDIA 8GB+ | Local voice, browser automation, dev servers |
| Power User | Workstation | 8+ cores | 32GB+ | NVIDIA 24GB+ VRAM | Full voice pipeline, local LLMs, parallel agents |
| Server | Cloud GPU | Any | 16GB+ | A100 / H100 | Production voice, multi-user, batch processing |
Cloud GPU providers for on-demand GPU access: NVIDIA Cloud, Vast.ai, RunPod, Lambda. See .agents/tools/infrastructure/cloud-gpu.md for the full deployment guide (SSH setup, Docker, model caching, cost optimization).
Note: Most aidevops features (infrastructure management, SEO, code quality, Git workflows) require no GPU. GPU is only needed for local AI model inference (voice pipeline, vision models, local LLMs).
# Install dependencies (auto-detected by setup.sh)
brew install sshpass jq curl mkcert dnsmasq fd ripgrep # macOS
sudo apt-get install sshpass jq curl dnsmasq fd-find ripgrep # Ubuntu/Debian
# Generate SSH key
ssh-keygen -t ed25519 -C "your-email@domain.com"
AI agents use fast file discovery tools for efficient codebase navigation:
| Tool | Purpose | Speed |
|---|---|---|
fd | Fast file finder (replaces find) | ~10x faster |
ripgrep | Fast content search (replaces grep) | ~10x faster |
Both tools respect .gitignore by default and are written in Rust for maximum performance.
Preference order for file discovery:
git ls-files '*.md' - Instant, git-tracked files onlyfd -e md - Fast, respects .gitignorerg --files -g '*.md' - Fast, respects .gitignoreThe setup script offers to install these tools automatically.
DESIGN.md standard: Canonical AI-readable design systems with YAML tokens, Markdown rationale, linting, Tailwind/DTCG export, and preview generation. aidevops keeps DESIGN.md as the source of truth for UI agents./design-artifact routes prototype, deck, email, poster, social carousel, and mobile mockup requests; /open-design manages optional Open Design companion workflows..od/ workspaces, and exports. It installs alongside aidevops only when requested; selected skills are ingested via aidevops build-agent optimisation, not imported verbatim.localdev-helper.sh can wrap Open Design or other dev servers with mkcert-backed .local routes when tools only expose localhost.workflows/ui-verification.md, email-design-test-helper.sh, design preview screenshots, and deck/media smoke tests provide evidence before generated artifacts ship..agents/tools/mobile/).serve-sim guidance and helpers cover local/mobile preview flows, prompt failure handling, and simulator smoke tests before mobile UI changes ship./design-artifact can route mobile mockups and app screens through the design stack, then hand implementation to mobile/testing specialists.Git CLI Enhancement Features:
aidevops secret) - subprocess injection + output redaction keeps secrets out of AI contextdocument-creation-helper.sh): Unified document format conversion, template-based creation, and OCR for scanned PDFs/images. Routes to the best available tool (pandoc, odfpy, LibreOffice, Tesseract, EasyOCR, GLM-OCR) based on format pair and availability. Supports 13+ formats (ODT, DOCX, PDF, MD, HTML, EPUB, PPTX, ODP, XLSX, ODS, RTF, CSV, TSV).PDF/OCR Tool Selection:
| Need | Tool | Why |
|---|---|---|
| Format conversion | Document Creation Agent | Auto-selects best tool, 13+ formats |
| Complex PDF to markdown | MinerU | Layout-aware, formulas, tables, 109-language OCR |
| Quick text extraction | GLM-OCR | Local, fast, no API keys, privacy-first |
| Structured JSON output | Unstract | Schema-based extraction, complex documents |
| Screen/window OCR | Peekaboo + GLM-OCR | peekaboo image --analyze --model ollama/glm-ocr |
| PDF text extraction | LibPDF | Native PDF parsing, no AI needed |
| Simple format conversion | Pandoc | Lightweight, broad format support |
| Scanned PDF OCR | Document Creation Agent | Auto-detects, routes to Tesseract/EasyOCR/GLM-OCR |
Quick start:
# Document creation agent
document-creation-helper.sh status # Check available tools
document-creation-helper.sh install --standard # Install core tools
document-creation-helper.sh convert report.pdf --to odt # Convert formats
document-creation-helper.sh convert scan.pdf --to md --ocr # OCR scanned PDF
document-creation-helper.sh template draft --type letter # Generate template
# GLM-OCR direct
ollama pull glm-ocr
ollama run glm-ocr "Extract all text" --images /path/to/document.png
See .agents/tools/ocr/glm-ocr.md for batch processing, PDF workflows, and Peekaboo integration.
matrix-dispatch-helper.sh)simplex-bot/)matterbridge-helper.sh)npx ctx7 setup --opencode --cli for docs lookup without MCP transport (useful fallback in shell-first workflows)local-model-helper.sh).local domains with HTTPS on port 443 (localdev-helper.sh)Model Context Protocol servers for real-time AI assistant integration. The framework configures these MCPs for OpenCode (TUI, Desktop, and Extension for Zed/VSCode).
MCP packages are installed globally via bun install -g for instant startup (no npx registry lookups). Run setup.sh or aidevops update-tools to update to latest versions.
| MCP | Purpose | Tier | API Key Required |
|---|---|---|---|
| Augment Context Engine | Semantic codebase retrieval | Global | Yes (Augment account) |
| Claude Code MCP | Claude as sub-agent | Global | No |
| Amazon Order History | Order data extraction | Per-agent | No |
| Chrome DevTools | Browser debugging & automation | Per-agent | No |
| Context7 | Library documentation lookup | Per-agent | No |
| Docker MCP | Container management | Per-agent | No |
| Google Analytics | Analytics data | Per-agent | Yes (Google API) |
| Google Search Console | Search performance data | Per-agent | Yes (Google API) |
| Grep by Vercel | GitHub code search | Per-agent | No |
| LocalWP | WordPress database access | Per-agent | No (local) |
| macOS Automator | macOS automation | Per-agent | No |
| Playwriter | Browser with extensions | Per-agent | No |
| QuickFile | Accounting API | Per-agent | Yes |
| Repomix | Codebase packing for AI context | Per-agent | No |
| Sentry | Error tracking | Per-agent | Yes |
| shadcn | UI component library | Per-agent | No |
| Socket | Dependency security | Per-agent | No |
| Unstract | Document data extraction | Per-agent | Yes |
| OpenAPI Search | Search and explore any OpenAPI spec | Per-agent | No |
| Cloudflare Code Mode | Full Cloudflare API (2,500+ endpoints via 2 tools) | Per-agent | Yes (Cloudflare) |
Tier explanation:
Performance optimization: MCP packages are installed globally via bun install -g for instant startup (~0.1s vs 2-3s with npx). The framework uses a three-tier loading strategy: MCPs load eagerly at startup or on-demand when their subagent is invoked. This reduces OpenCode startup time significantly.
These use direct API calls via curl, avoiding MCP server startup entirely:
| Integration | Purpose | API Key Required |
|---|---|---|
| Ahrefs | SEO analysis & backlinks | Yes |
| DataForSEO | SERP, keywords, backlinks, on-page | Yes |
| Serper | Google Search API (web, images, news) | Yes |
| Semrush | Domain analytics, keywords, backlinks, competitor research | Yes |
| ContentKing | Real-time SEO monitoring, change tracking, issues | Yes |
| WebPageTest | Real-world performance testing from 40+ global locations | Yes |
| Hostinger | Hosting management | Yes |
| NeuronWriter | Content optimization & NLP analysis | Yes |
| Outscraper | Google Maps & business data extraction | Yes |
Context & Codebase:
Browser Automation (browser tool suite + anti-detect stack, benchmarked):
/auto-browse orchestrates the tools below to learn messy browser tasks, choose the cheapest reliable path, preserve private profile/session state under ~/.aidevops/, and graduate reusable private agents or sanitized todo/ plans--session isolation, built-in tracing--session parallel, ref-based element targeting, iOS Simulator support (macOS)arun_many parallel (1.7x), LLM-ready markdownMobile & Simulator Testing:
SEO & Research:
Data Extraction:
Performance & Security:
ip-reputation-helper.sh)WordPress & Development:
CRM & Marketing:
Accounts & Finance:
Document Processing & OCR:
# Install all MCP integrations
bash .agents/scripts/setup-mcp-integrations.sh all
# Install specific integration
bash .agents/scripts/setup-mcp-integrations.sh stagehand # JavaScript version
bash .agents/scripts/setup-mcp-integrations.sh stagehand-python # Python version
bash .agents/scripts/setup-mcp-integrations.sh stagehand-both # Both versions
bash .agents/scripts/setup-mcp-integrations.sh chrome-devtools
OpenCode includes built-in LSP servers for 35+ languages. For aidevops projects that use Markdown and TOON extensively, add these optional LSP servers to your opencode.json for real-time diagnostics during editing:
{
"lsp": {
"markdownlint": {
"command": ["markdownlint-language-server", "--stdio"],
"extensions": [".md"]
},
"toon-lsp": {
"command": ["toon-lsp"],
"extensions": [".toon"]
}
}
}
Install the servers:
npm install -g markdownlint-language-server # Markdown diagnostics
cargo install toon-lsp # TOON syntax validation
These catch formatting and syntax issues during editing, reducing preflight/postflight fix cycles.
Browser tool suite + anti-detect stack + device emulation, benchmarked and integrated for AI-assisted web automation, dev testing, mobile/responsive testing, data extraction, and bot detection evasion. Agents automatically select the optimal tool based on task requirements.
For repeatable browser operations and web data mining, use /auto-browse. It runs an intake and learning loop that starts with cheap fetch/API/crawler options, escalates to deterministic or high-agency browser tools only when needed, and then graduates the workflow into a private custom agent, helper, schema, or sanitized todo/ plan. Account-specific workflows, cookies, profile state, downloads, and traces stay in private aidevops user data; only generalized plans should be committed to the repo.
Tested on macOS ARM64, all headless, warm daemon:
| Test | Playwright | playwright-cli | dev-browser | agent-browser | Crawl4AI | Playwriter | Stagehand | browser-use |
|---|---|---|---|---|---|---|---|---|
| Navigate + Screenshot | 1.43s | ~1.9s | 1.39s | 1.90s | 2.78s | 2.95s | 7.72s | Agentic; not in local deterministic suite |
| Form Fill (4 fields) | 0.90s | ~1.4s | 1.34s | 1.37s | N/A | 2.24s | 2.58s | Agentic; use for fuzzy forms |
| Data Extraction (5 items) | 1.33s | ~1.5s | 1.08s | 1.53s | 2.53s | 2.68s | 3.48s | Agentic; compress to deterministic when stable |
| Multi-step (click + nav) | 1.49s | ~2.0s | 1.49s | 3.06s | N/A | 4.37s | 4.48s | Upstream BU Bench covers 100 real-world tasks |
| Parallel (3 sessions) | 1.6s | ~2.0s | N/A | 2.0s | 3.0s | N/A | Slow | Prefer Browser Use Cloud for scale |
| Feature | Playwright | playwright-cli | dev-browser | agent-browser | Crawl4AI | Playwriter | Stagehand | browser-use |
|---|---|---|---|---|---|---|---|---|
| Headless | Yes | Yes (default) | Yes | Yes (default) | Yes | No (your browser) | Yes | Yes |
| Proxy/VPN | Full | No | Via args | No | Full | Your browser | Via args | Cloud/profiles |
| Extensions | Yes (persistent) | No | Yes (profile) | No | No | Yes (yours) | Possible | Profile/browser dependent |
| Password managers | Partial (needs unlock) | No | Partial | No | No | Yes (unlocked) | No | Profile/browser dependent |
| Device emulation | Full (100+ devices) | No | No | No | No | No | Via Playwright | Profile/browser dependent |
| Parallel sessions | 5 ctx/2.1s | --session | Shared | 3 sess/2.0s | arun_many 1.7x | Shared | Per-instance | Cloud for scale |
| Session persistence | storageState | Profile dir | Profile dir | state save/load | user_data_dir | Your browser | Per-instance | BrowserProfile / Cloud |
| Tracing | Full API | Built-in CLI | Via Playwright | Via Playwright | No | Via CDP | Via Playwright | History + Browser Harness CLI |
| Natural language | No | No | No | No | LLM extraction | No | Yes | Yes |
| Self-healing | No | No | No | No | No | No | Yes | Yes |
| Agent skill | No | No | No | No | No | No | No | browser-use skill |
| iOS Simulator | No | No | No | Yes (macOS) | No | No | No | No |
| Maintainer | Microsoft | Microsoft | Community | Vercel | Community | Community | Browserbase | Browser Use |
| Need | Tool | Why |
|---|---|---|
| Repeatable browser workflow | /auto-browse | Learns, optimizes, preserves private profile state, and graduates reusable workflows |
| Fastest automation | Playwright | 0.9s form fill, parallel contexts |
| AI agent (CLI) | playwright-cli | Microsoft official, --session isolation, built-in tracing |
| Stay logged in | dev-browser | Profile persists across restarts |
| Your extensions/passwords | Playwriter | Already unlocked in your browser |
| Bulk extraction | Crawl4AI | Purpose-built, parallel, LLM-ready output |
| Self-hosted crawling | WaterCrawl | Docker deployment, web search, sitemap generation |
| CLI/CI/CD | playwright-cli or agent-browser | No server needed, --session isolation |
| iOS mobile testing | agent-browser | Real Safari in iOS Simulator (macOS only) |
| Apple Simulator streaming | serve-sim | Browser-visible iOS/iPad/watch simulator stream and control (macOS Apple Silicon) |
| Unknown pages | Stagehand | Natural language, self-healing |
| High-agency browser tasks | browser-use | AI-native browser agent, CLI 3.0, Browser Harness, cloud/stealth/CAPTCHA options |
| Performance debugging | Chrome DevTools MCP | Companion tool, pairs with any browser |
| Mobile/tablet emulation | Playwright | 100+ device presets, viewport, touch, geolocation, locale |
| Authenticated one-off scrape | curl-copy | DevTools "Copy as cURL" → paste to terminal/AI |
| Bot detection evasion | Anti-detect stack | Camoufox (full) or rebrowser-patches (quick) |
| Multi-account | Browser profiles | Persistent fingerprint + proxy per account |
Agents use lightweight methods instead of expensive vision API calls:
| Method | Speed | Token Cost | Use For |
|---|---|---|---|
| ARIA snapshot | ~0.01s | 50-200 tokens | Forms, navigation, interactive elements |
| Text extraction | ~0.002s | Text length | Reading content |
| Element scan | ~0.002s | ~20/element | Form filling, clicking |
| Screenshot | ~0.05s | ~1K tokens (vision) | Visual debugging only |
See .agents/tools/browser/browser-automation.md for the full decision tree, browser-use.md for Browser Use CLI 3.0 guidance, and browser-benchmark.md for reproducible benchmark scripts.
Test responsive layouts and mobile-specific behavior using Playwright's built-in device emulation. Supports 100+ device presets with viewport, user agent, touch events, device scale factor, geolocation, locale/timezone, permissions, color scheme, offline mode, and network throttling.
Common device presets:
| Device | Viewport | Scale | Touch |
|---|---|---|---|
iPhone 15 | 393x852 | 3 | Yes |
iPad Pro 11 | 834x1194 | 2 | Yes |
Pixel 7 | 412x915 | 2.625 | Yes |
Galaxy S9+ | 320x658 | 4.5 | Yes |
Desktop Chrome | 1280x720 | 1 | No |
Emulation capabilities:
| Feature | Example |
|---|---|
| Device presets | devices['iPhone 13'] - viewport, UA, touch, scale |
| Viewport/HiDPI | viewport: { width: 2560, height: 1440 }, deviceScaleFactor: 2 |
| Geolocation | geolocation: { longitude: -74.006, latitude: 40.7128 } |
| Locale/timezone | locale: 'de-DE', timezoneId: 'Europe/Berlin' |
| Color scheme | colorScheme: 'dark' |
| Offline mode | offline: true |
| Permissions | permissions: ['geolocation', 'notifications'] |
| Network throttling | CDP-based Slow 3G / Fast 3G emulation |
Recipes included: Responsive breakpoint testing, multi-device parallel testing, touch gesture testing, geolocation-dependent features, dark mode visual regression, and network condition emulation.
See .agents/tools/browser/playwright-emulation.md for complete documentation with code examples.
Open-source alternative to AdsPower, GoLogin, and OctoBrowser for multi-account automation and bot detection evasion.
Architecture:
Layer 4: CAPTCHA Solving → CapSolver (existing)
Layer 3: Network Identity → Proxies (residential/SOCKS5/VPN per profile)
Layer 2: Browser Identity → Camoufox (C++ fingerprint injection)
Layer 1: Automation Stealth → rebrowser-patches (CDP leak prevention)
Layer 0: Browser Engine → Playwright (existing)
Profile Types:
| Type | Cookies | Fingerprint | Use Case |
|---|---|---|---|
| Persistent | Saved | Fixed per profile | Account management, stay logged in |
| Clean | None | Random each launch | Scraping, one-off tasks |
| Warm | Saved | Fixed | Pre-warmed accounts (browsing history) |
| Disposable | None | Random | Single-use, maximum anonymity |
Quick Start:
# Setup
anti-detect-helper.sh setup
# Create profile with proxy
anti-detect-helper.sh profile create "my-account" --type persistent --os macos
# Launch (Camoufox with auto-generated fingerprint)
anti-detect-helper.sh launch --profile "my-account" --headless
# Test detection (BrowserScan, SannyBot)
anti-detect-helper.sh test --profile "my-account"
# Warm up profile with browsing history
anti-detect-helper.sh warmup "my-account" --duration 30m
Engine Selection:
| Engine | Stealth Level | Speed | Best For |
|---|---|---|---|
| Camoufox (Firefox) | High (C++ level) | Medium | Full anti-detect, fingerprint rotation |
| rebrowser-patches (Chromium) | Medium (CDP patches) | Fast | Quick stealth on existing Playwright code |
See .agents/tools/browser/anti-detect-browser.md for the full decision tree and subagent index.
Repomix packages your codebase into AI-friendly formats for sharing with AI assistants. This framework includes optimized Repomix configuration for consistent context generation.
| Use Case | Tool | When to Use |
|---|---|---|
| Interactive coding | Augment Context Engine | Real-time semantic search during development |
| Share with external AI | Repomix | Self-contained snapshot for ChatGPT, Claude web, etc. |
| Architecture review | Repomix (compress) | 80% token reduction, structure only |
| CI/CD integration | GitHub Action | Automated context in releases |
# Pack current repo with configured defaults
npx repomix
# Compress mode (~80% smaller, structure only)
npx repomix --compress
# Or use the helper script
.agents/scripts/context-builder-helper.sh pack # Full context
.agents/scripts/context-builder-helper.sh compress # Compressed
| File | Purpose |
|---|---|
repomix.config.json | Default settings (style, includes, security) |
.repomixignore | Additional exclusions beyond .gitignore |
repomix-instruction.md | Custom AI instructions included in output |
.agents/ contentRepomix runs as an MCP server for direct AI assistant integration:
{
"repomix": {
"type": "local",
"command": ["repomix", "--mcp"],
"enabled": true
}
}
Install globally first:
bun install -g repomix(done automatically bysetup.sh)
See .agents/tools/context/context-builder.md for complete documentation.
Augment Context Engine provides semantic codebase retrieval - understanding your code at a deeper level than simple text search. It's the recommended tool for real-time interactive coding sessions.
| Feature | grep/glob | Augment Context Engine |
|---|---|---|
| Text matching | Exact patterns | Semantic understanding |
| Cross-file context | Manual | Automatic |
| Code relationships | None | Understands dependencies |
| Natural language | No | Yes |
Use it to:
# 1. Install Auggie CLI (requires Node.js 22+)
npm install -g @augmentcode/auggie@prerelease
# 2. Authenticate (opens browser)
auggie login
# 3. Verify installation
auggie token print
Add to your AI assistant's MCP configuration:
OpenCode (~/.config/opencode/opencode.json):
{
"mcp": {
"augment-context-engine": {
"type": "local",
"command": ["auggie", "--mcp"],
"enabled": true
}
}
}
claude-code CLI:
claude mcp add-json auggie-mcp --scope user '{"type":"stdio","command":"auggie","args":["--mcp"]}'
Test with this prompt:
What is this project? Please use codebase retrieval tool to get the answer.
The AI should provide a semantic understanding of your project architecture.
| Use Case | Tool | When to Use |
|---|---|---|
| Interactive coding | Augment Context Engine | Real-time semantic search during development |
| Share with external AI | Repomix | Self-contained snapshot for ChatGPT, Claude web, etc. |
| Architecture review | Repomix (compress) | 80% token reduction, structure only |
| CI/CD integration | Repomix GitHub Action | Automated context in releases |
See .agents/tools/context/augment-context-engine.md for complete documentation.
llm-tldr extracts code structure and semantics, saving ~95% tokens compared to raw code. From the Continuous-Claude project.
# Install
pip install llm-tldr
# CLI usage
tldr tree ./src # File structure with line counts
tldr structure src/auth.py # Code skeleton (classes, functions)
tldr context src/auth.py # Full semantic analysis
tldr search "authentication" ./src # Semantic code search
tldr impact src/auth.py validate # What would change affect?
MCP Integration:
{
"llm-tldr": {
"command": "tldr-mcp",
"args": ["--project", "${workspaceFolder}"]
}
}
| Feature | Token Savings | Use Case |
|---|---|---|
| Structure extraction | 90% | Understanding code layout |
| Context analysis | 95% | Comprehensive code understanding |
| Semantic search | N/A | Finding code by meaning |
| Impact analysis | N/A | Change risk assessment |
See .agents/tools/context/llm-tldr.md for complete documentation.
aidevops implements the Agent Skills standard for cross-tool compatibility. Skills are auto-discovered by compatible AI assistants.
Generated SKILL.md files in ~/.aidevops/agents/ provide skill metadata following the Agent Skills standard. These are discoverable by any compatible tool.
aidevops is registered as a Claude Code plugin marketplace. Install with two commands:
/plugin marketplace add marcusquinn/aidevops
/plugin install aidevops@aidevops
This installs the complete framework: 14 primary agents, 2,150+ subagents, and 1,630+ helper scripts.
Import skills from GitHub or ClawdHub using the aidevops skill CLI:
# Import from GitHub (auto-detects format)
aidevops skill add owner/repo
# Import from ClawdHub (skill registry with vector search)
aidevops skill add clawdhub:caldav-calendar
aidevops skill add https://clawdhub.com/owner/slug
# More examples
aidevops skill add anthropics/skills/pdf # Specific skill from multi-skill repo
aidevops skill add vercel-labs/agent-skills # All skills from a repo
aidevops skill add expo/skills --name expo-dev # Custom name
aidevops skill add owner/repo --dry-run # Preview without changes
Supported sources:
owner/repo or full URL) — fetched via git cloneclawdhub:slug or full URL) — fetched via Playwright browser automationSupported formats:
SKILL.md - Agent Skills standard (preferred)AGENTS.md - Claude Code agents format.cursorrules - Cursor rules format (auto-converted)Features:
aidevops skill check).agents/configs/skill-sources.jsonHow it differs from npx add-skill:
aidevops skill add | npx add-skill | |
|---|---|---|
| Target | Converts to aidevops format in .agents/ | Copies SKILL.md to agent-specific dirs |
| Tracking | Git commit-based upstream tracking | Lock file with content hashes |
| Telemetry | Disabled | Sends anonymous install counts |
| Scope | OpenCode-first | 22+ agents |
| Updates | aidevops skill check (GitHub API) | npx skills check (Vercel API) |
See .agents/scripts/add-skill-helper.sh for implementation details.
Agents are specialized AI personas with focused knowledge and tool access. Instead of giving your AI assistant access to everything at once (which wastes context tokens), agents provide targeted capabilities for specific tasks.
Call them in your AI assistant conversation with a simple @mention
| Concept | Description |
|---|---|
| Main Agent | Domain-focused assistant (e.g., WordPress, SEO, DevOps) |
| Subagent | Specialized assistant for specific services (invoked with @mention) |
| MCP Tools | Only loaded when relevant agent is invoked (saves tokens) |
Primary agents live at .agents/<name>.md. Each is a domain expert with its own system prompt, tool permissions, and subagent roster. MCPs are loaded on-demand per subagent, not per primary agent.
| Name | File | Purpose | Model Tier |
|---|---|---|---|
| Build+ | build-plus.md | Code: features, bug fixes, refactors, CI, full-loop delivery (default) | thinking |
| Automate | automate.md | Scheduling, dispatch, monitoring, background orchestration | standard |
| Aidevops | aidevops.md | Framework development — meta-agent for improving aidevops itself | thinking |
| Business | business.md | Company orchestration, financial ops, invoicing, strategy | standard |
| Content | content.md | Content creation across blog, video, audio, image, social | thinking |
| Health | health.md | Health and wellness content, fitness, nutrition | thinking |
| Legal | legal.md | Legal compliance, terms, privacy, GDPR | thinking |
| Marketing-Sales | marketing-sales.md | Email campaigns, CRM, outreach, paid ads, direct response, CRO | thinking |
| Product | product.md | Product management, PRDs, roadmaps, requirements capture | thinking |
| Research | research.md | Technical and market research, competitive analysis | standard |
| SEO | seo.md | SEO audits, keyword research, GSC, schema, technical SEO | thinking |
Specialist subagents (e.g. @wordpress, @seo, Build-Agent, Build-MCP, etc.) live under .agents/tools/ or as mode: subagent files and are invoked via @mention when domain expertise is needed. See subagent-index.toon for the full roster. For read-only macOS Activity Monitor and background-efficiency audits, use @macos-activity-cleaner or /aidevops-macos-activity-cleaner.
| Client | Invocation |
|---|---|
| OpenCode | Tab through the agent picker in the UI — Build+ is the default. Main agents are registered as top-level agents in OpenCode's config. |
| Claude Code / Codex / Cursor / Droid / Kiro / Continue / Kimi / Qwen / Amp / Windsurf / Gemini CLI | Slash command, namespaced with the aidevops- prefix: /aidevops-build-plus, /aidevops-automate, /aidevops-seo, etc. |
| Aider | No native slash command support — use a shell alias: alias aider-build='aider --message-file ~/.aidevops/agents/build-plus.md'. |
The aidevops- prefix differentiates framework commands from each client's native slash commands and groups them alphabetically in the command picker. It applies to every aidevops slash command — not just main agents — so /aidevops-preflight, /aidevops-release, /aidevops-commit etc. all sort together in your / menu.
The framework installs itself across these clients. Slash commands, agent definitions, and (optionally) session-memory mining are wired up per-client by setup.sh, gated on per-client feature flags in .agents/scripts/runtime-registry.sh.
| Client | Slash commands | Agent dir | Memory mining | Notes |
|---|---|---|---|---|
| OpenCode | ✅ ~/.config/opencode/command/ | config-based | ✅ default | Native tab-through primary agents |
| Claude Code | ✅ ~/.claude/commands/ | ✅ ~/.claude/agents/ | ✅ default | Full feature parity |
| Codex CLI | ✅ ~/.codex/prompts/ | — | ✅ default | Invoked as /prompts:aidevops-<name> |
| Cursor | ✅ ~/.cursor/commands/ (≥1.6) | ✅ ~/.cursor/agents/ | ✅ default | Frontmatter stripped (not supported) |
| Droid (Factory) | ✅ ~/.factory/commands/ | — | opt-in | — |
| Gemini CLI | ✅ ~/.gemini/commands/ | — | opt-in | Converted to TOML (prompt = """...""") |
| Kimi CLI | ✅ ~/.kimi/skills/<name>/SKILL.md | ✅ ~/.kimi/agents/ | opt-in | Directory-per-skill + auto name: matching |
| Qwen Code | ✅ ~/.qwen/commands/ | ✅ ~/.qwen/agents/ | opt-in | Sub-agents + skills |
| Continue | ✅ ~/.continue/prompts/ | — | opt-in | .prompt ext + invokable: true |
| Kiro | ✅ ~/.kiro/steering/ | — | opt-in | inclusion: manual for slash access |
| Kilo Code | custom modes | — | opt-in | Uses modes instead of commands |
| Windsurf | repo-local .windsurf/workflows/ | — | ❌ (protobuf) | Symlinked by aidevops init |
| Amp (Sourcegraph) | repo-local .agents/commands/ | ✅ ~/.amp/agents/ | ❌ (cloud) | Path match is native |
| Aider | shell alias workaround | — | ❌ (per-repo md) | Native custom commands open upstream |
Feature flags per client (agents / commands / memory) live in runtime-registry.sh and can be overridden at install time via environment variables:
AIDEVOPS_FEATURE_MEMORY_CLAUDE_CODE=no setup.sh
AIDEVOPS_FEATURE_COMMANDS_CURSOR=no setup.sh
Memory mining defaults are deliberately conservative: only OpenCode, Claude Code, Codex, and Cursor are opted in by default. All other clients default to off — enable them case-by-case after reviewing what the mining job will read.
The long-term direction is to make slash commands and @mentions unnecessary altogether. A progressive disclosure layer should load the right domain agents and tools into context based on the nature of the conversation — you should never have to remember agent names or prefix commands. The aidevops- slash commands documented above are a stepping stone: they standardise routing across every client we support, and will eventually be auto-invoked by the router rather than typed by hand.
These are examples of subagents that have supporting MCPs enabled. See .agents/ for the full list of 2,150+ subagents organized by domain.
| Agent | Purpose | MCPs Enabled |
|---|---|---|
@hostinger | Hosting, WordPress, DNS, domains | hostinger-api |
@hetzner | Cloud servers, firewalls, volumes | hetzner-* (multi-account) |
@wordpress | Local dev, MainWP management | localwp, context7 |
@seo | Search Console, keyword research, domain intelligence | gsc, ahrefs, serper, context7 |
@dataforseo | SERP, keywords, backlinks, on-page analysis | (curl subagent) |
@domain-research | DNS intelligence: rDNS, subdomains, CNAMEs (THC + Reconeer) | (API-based) |
@serper | Google Search API (web, images, news, places) | serper |
@list-keys | List all configured API keys and storage locations | (read-only) |
@code-standards | Quality standards reference, compliance checking | context7 |
@browser-automation | Testing, scraping, DevTools | chrome-devtools, context7 |
@performance | Core Web Vitals, network analysis, accessibility | chrome-devtools |
@git-platforms | GitHub, GitLab, Gitea | context7 |
@sentry | Error monitoring, Next.js SDK setup | sentry |
@socket | Dependency security scanning | socket |
@security-analysis | AI-powered vulnerability detection (OSV, Ferret, git history) | osv-scanner, gemini-cli-security |
@secretlint | Detect exposed secrets in code | (Docker-based) |
@snyk | Security vulnerability scanning | (API-based) |
@auditing | Code auditing services and security analysis | (API-based) |
@agent-review | Session analysis, agent improvement (under build-agent/) | (read/write only) |
# Install aidevops agents for OpenCode
.agents/scripts/generate-opencode-agents.sh
# Check status
.agents/scripts/generate-opencode-agents.sh # Shows status after generation
aidevops is available in the Claude marketplace:
/plugin marketplace add marcusquinn/aidevops
/plugin install aidevops-all@aidevops
Agent Skills (SKILL.md): Auto-discovered from ~/.aidevops/agents/ after running setup.sh. Compatible with any tool that supports the Agent Skills standard.
End every session by calling @agent-review to analyze what worked and what didn't:
@agent-review analyze this session and suggest improvements to the agents used
The review agent will:
This creates a feedback loop:
Session → @agent-review → Improvements → Better Agents → Better Sessions
↓
PR to aidevops repo (optional)
Contributing improvements:
@agent-review create a PR for improvement #2
The agent will create a safe linked worktree, apply changes, and submit a PR to marcusquinn/aidevops with a structured description. Your real-world usage helps improve the framework for everyone.
Code quality learning loop:
The @code-quality agent also learns from issues. After fixing violations from SonarCloud, Codacy, ShellCheck, etc., it analyzes patterns and updates framework guidance to prevent recurrence:
Quality Issue → Fix Applied → Pattern Identified → Framework Updated → Issue Prevented
Slash commands provide quick access to common workflows directly from the OpenCode prompt. Type / to see available commands.
Planning & Task Management:
| Command | Purpose |
|---|---|
/list-todo | List tasks with sorting, filtering, and grouping |
/save-todo | Save discussion as task or plan (auto-detects complexity) |
/plan-status | Check status of plans in TODO.md and todo/PLANS.md |
/create-prd | Create a Product Requirements Document for complex features |
/generate-tasks | Generate implementation tasks from a PRD |
/log-time-spent | Log time spent on a task for tracking |
/ready | Show tasks with no open blockers (Beads integration) |
/sync-beads | Sync TODO.md/PLANS.md with Beads task graph |
/remember | Store knowledge for cross-session recall |
/recall | Search memories from previous sessions |
Plans are tracked in TODO.md (all tasks) and todo/PLANS.md (complex execution plans). Task dependencies are visualized with Beads.
/list-todo options:
| Option | Example | Purpose |
|---|---|---|
--priority | /list-todo -p | Sort by priority (high → low) |
--estimate | /list-todo -e | Sort by time estimate (shortest first) |
--tag | /list-todo -t seo | Filter by tag |
--owner | /list-todo -o marcus | Filter by assignee |
--estimate | /list-todo --estimate "<2h" | Filter by estimate range |
--group-by | /list-todo -g tag | Group by tag, owner, status, or estimate |
--plans | /list-todo --plans | Include full plan details |
--compact | /list-todo --compact | One-line per task |
Time Tracking: Tasks support time estimates as active session time (excluding AFK gaps) with the format ~4h started:2025-01-15T10:30Z. The session-time-helper.sh analyses real session data to calibrate estimates vs actuals.
Risk Levels: Tasks support risk:low/med/high indicating human oversight needed:
| Risk | Oversight | Example |
|---|---|---|
low | Autonomous | Docs, formatting, simple refactors |
med | Supervised | Feature implementation, API changes |
high | Engaged | Security, data migrations, infrastructure |
Configure time tracking per-repo via .aidevops.json.
Development Workflow (typical order):
| Command | Purpose |
|---|---|
/context | Build AI context with Repomix for complex tasks |
/feature | Start a new feature linked-worktree workflow |
/bugfix | Start a bugfix linked-worktree workflow |
/hotfix | Start an urgent hotfix workflow |
/linters-local | Run local linting (shfmt, ShellCheck, secretlint) |
/code-audit-remote | Run remote auditing (CodeRabbit, Codacy, SonarCloud) |
/code-standards | Check against documented quality standards |
/code-simplifier | Simplify and refine code for clarity and maintainability |
/testing-setup | Interactive per-repo testing infrastructure setup with bundle-aware defaults |
/list-keys | List all configured API keys and their storage locations |
/performance | Web performance audit (Core Web Vitals, Lighthouse, PageSpeed) |
/pr | Unified PR workflow (orchestrates all checks) |
/cross-review | Dispatch prompt to multiple AI models, diff results, auto-score |
/tech-stack | Detect technology stacks of URLs or find sites using specific technologies |
/mission | Scope a high-level goal into milestones and features for autonomous execution |
/budget-analysis | Analyze AI model spend, burn rate, and cost optimization opportunities |
/report-token-use | Generate a local per-session token, model, compaction, and MCP-use report |
Content Workflow:
| Command | Purpose |
|---|---|
/humanise | Remove AI writing patterns, make text sound human |
Media:
| Command | Purpose |
|---|---|
/yt-dlp | Download YouTube videos/audio/playlists, extract transcripts, convert local files |
SEO Workflow:
| Command | Purpose |
|---|---|
/keyword-research | Seed keyword expansion with volume, CPC, difficulty |
/autocomplete-research | Google autocomplete long-tail discovery |
/keyword-research-extended | Full SERP analysis with weakness detection |
/webmaster-keywords | Keywords from GSC + Bing for your verified sites |
/neuronwriter | Content optimization with NLP term recommendations and scoring |
/seo-export | Export SEO data from GSC, Bing, Ahrefs, DataForSEO to TOON format |
/seo-analyze | Analyze exported data for quick wins, striking distance, low CTR |
/seo-opportunities | Combined export + analysis workflow |
/seo-audit | Comprehensive SEO audit: technical, on-page, content, E-E-A-T |
SEO Debugging & Auditing (subagents in seo/):
| Subagent | Purpose |
|---|---|
@seo-audit | Comprehensive SEO audit: technical, on-page, content quality, E-E-A-T |
@debug-opengraph | Validate Open Graph meta tags, preview social sharing |
@debug-favicon | Validate favicon setup across platforms (ico, apple-touch, manifest) |
Email Deliverability:
| Command | Purpose |
|---|---|
/email-health-check | Check SPF/DKIM/DMARC/MX records and blacklist status |
Release Workflow (in order):
| Command | Purpose |
|---|---|
/preflight | Run quality checks before release |
/changelog | Update CHANGELOG.md with recent changes |
/version-bump | Bump version following semver |
/release | Full release workflow (bump, tag, GitHub release, auto-changelog) |
/postflight | Verify release health after deployment |
Auto-Task Completion: The release workflow automatically marks tasks as complete when commit messages reference them (e.g., Closes t037, Fixes t042). Tasks in TODO.md are updated with completed: timestamps.
Documentation:
| Command | Purpose |
|---|---|
/readme | Create or update README.md (supports --sections for partial updates) |
Meta/Improvement:
| Command | Purpose |
|---|---|
/add-skill | Import external skills from GitHub repos (SKILL.md, AGENTS.md, .cursorrules) |
/agent-review | Analyze session and suggest agent improvements |
/session-analysis | Find evidence-backed harness, model-routing, and repository optimisation opportunities |
/session-review | Review session for completeness and capture learnings |
/full-loop | End-to-end development loop (task → preflight → PR → postflight → deploy) |
/preflight-loop | Run preflight checks iteratively until all pass |
/runners | Batch dispatch tasks to parallel agents (task IDs, PR URLs, or descriptions) |
/log-issue-aidevops | Report issues with aidevops (gathers diagnostics, checks duplicates, creates GitHub issue) |
AI Model Comparison:
| Command | Purpose |
|---|---|
/compare-models | Compare AI models by pricing, context, capabilities (with live web data) |
/compare-models-free | Compare AI models using offline embedded data only (no web fetches) |
/score-responses | Score and compare actual model responses with structured criteria |
/route | Suggest optimal model tier for a task description |
The Ralph Loop (named after Ralph Wiggum's persistent optimism) enables autonomous iterative development. The AI keeps working on a task until it's complete, automatically resolving issues that arise.
How it works:
Task → Implement → Check → Fix Issues → Re-check → ... → Complete
↑ ↓
└────────────────────┘ (loop until done)
Usage:
# Run quality checks iteratively until all pass
/preflight-loop --auto-fix --max-iterations <MAX_ITERATIONS>
# Or run changed-file checks directly (the default)
.agents/scripts/linters-local.sh
# Run every broad gate at a release boundary
.agents/scripts/linters-local.sh --full
Note: Store any API credentials securely via environment variables or .env files (never commit credentials to version control).
Key features:
See .agents/workflows/ralph-loop.md for the full workflow guide.
The Full Loop chains all development phases into a single automated workflow:
Task Development → Preflight → PR Create → PR Review → Postflight → Deploy
Usage:
# Start a full development loop
/full-loop "Implement feature X with tests"
# With options
/full-loop "Fix bug Y" --max-task-iterations 30 --skip-postflight
Options:
| Option | Description |
|---|---|
--max-task-iterations N | Max iterations for task (default: 50) |
--skip-preflight | Skip preflight checks |
--skip-postflight | Skip postflight monitoring |
--no-auto-pr | Pause for manual PR creation |
The loop pauses for human input at merge approval, rollback decisions, and scope changes.
See .agents/scripts/commands/full-loop.md for complete documentation.
Work on multiple branches simultaneously without stashing or switching. Each branch gets its own directory.
Recommended: Worktrunk (wt) - Git worktree management with shell integration, CI status, and PR links:
# Install (macOS/Linux)
brew install max-sixty/worktrunk/wt && wt config shell install
# Restart your shell for shell integration to take effect
# Create worktree + cd into it
wt switch -c feature/my-feature
# Create worktree + start any AI CLI (-x runs command after switch)
wt switch -c -x claude feature/ai-task
# List worktrees with CI status and PR links
wt list
# Merge + cleanup (squash/rebase options)
wt merge
Fallback (no dependencies):
~/.aidevops/agents/scripts/worktree-helper.sh add feature/my-feature
# Creates: ~/Git/{repo}-feature-my-feature/ (cd there manually)
~/.aidevops/agents/scripts/worktree-helper.sh list
~/.aidevops/agents/scripts/worktree-helper.sh clean
Benefits:
Worktree-first workflow: The pre-edit check now enforces worktrees as the default when creating branches, keeping your main directory on main. This prevents uncommitted changes from blocking branch switches and ensures parallel sessions don't inherit wrong branch state.
See .agents/workflows/worktree.md for the complete guide and .agents/tools/git/worktrunk.md for Worktrunk documentation.
Spawn new AI sessions for parallel work or fresh context. The framework detects natural session completion points and suggests next steps.
Completion signals:
| Signal | Meaning |
|---|---|
| PR merged | Branch work complete |
| Release published | Version shipped |
| All tests passing | Quality gates satisfied |
| User says "done" | Explicit completion |
Spawning options:
# Background session (same terminal)
opencode --non-interactive --prompt "Continue with feature X" &
# New terminal tab (macOS)
osascript -e 'tell application "Terminal" to do script "cd ~/Git/project && opencode"'
# Worktree-based (isolated branch) - recommended
wt switch -c -x opencode feature/next-feature # Worktrunk: create + start AI CLI
# Or fallback:
# ~/.aidevops/agents/scripts/worktree-helper.sh add feature/next-feature
# cd ~/Git/{repo}-feature-next-feature && opencode
Session handoff pattern:
When ending a session, the AI provides a continuation prompt for the next session:
## Continuation Prompt
[Copy this to start a new session with full context]
See .agents/workflows/session-manager.md for the complete guide.
"Compound, then clear" - Sessions should build on each other. The memory system stores knowledge, patterns, and learnings for future sessions using SQLite FTS5 for fast full-text search, with opt-in semantic similarity search via vector embeddings.
Slash commands:
| Command | Purpose |
|---|---|
/remember {content} | Store a memory with AI-assisted categorization |
/recall {query} | Search memories by keyword |
/recall --recent | Show 10 most recent memories |
/recall --auto-only | Search only auto-captured memories |
/recall --stats | Show memory statistics |
/memory-log | Show recent auto-captured memories |
/patterns {task} | Show success/failure patterns for a task type |
/route {task} | Suggest optimal model tier for a task |
Memory types:
| Type | Use For |
|---|---|
WORKING_SOLUTION | Fixes that worked |
FAILED_APPROACH | What didn't work (avoid repeating) |
CODEBASE_PATTERN | Project conventions |
USER_PREFERENCE | Developer preferences |
TOOL_CONFIG | Tool setup notes |
DECISION | Process, workflow, or policy choices (naming, release cadence, branching) |
CONTEXT | Background info |
ARCHITECTURAL_DECISION | System-level architecture (service boundaries, data flow, tech stack) |
ERROR_FIX | Bug fixes and patches |
OPEN_THREAD | Unresolved questions or follow-ups |
SUCCESS_PATTERN | Approaches that consistently work |
FAILURE_PATTERN | Approaches that consistently fail |
Semantic search (opt-in):
# Enable semantic similarity search (~90MB model download)
memory-embeddings-helper.sh setup
# Search by meaning, not just keywords
memory-helper.sh recall "how to optimize queries" --semantic
Pattern tracking:
# Record what worked (via memory system)
memory-helper.sh store --type SUCCESS_PATTERN --content "Structured debugging found root cause" --tags "bugfix,standard"
# Record what failed
memory-helper.sh store --type FAILURE_PATTERN --content "Blind refactor without tests caused regressions" --tags "refactor"
# Get suggestions for a new task
memory-helper.sh recall "refactor auth middleware" --semantic
Auto-capture: AI agents automatically store memories using --auto flag when they detect working solutions, failed approaches, or decisions. Privacy filters strip <private> tags and reject secret patterns.
CLI usage:
# Store a memory
~/.aidevops/agents/scripts/memory-helper.sh store --type "WORKING_SOLUTION" --content "Fixed CORS with nginx headers" --tags "cors,nginx"
# Store auto-captured memory (from AI agent)
~/.aidevops/agents/scripts/memory-helper.sh store --auto --content "Fixed CORS with nginx headers" --type WORKING_SOLUTION
# Recall memories (keyword search - default)
~/.aidevops/agents/scripts/memory-helper.sh recall "cors"
# Recall memories (semantic similarity - opt-in)
~/.aidevops/agents/scripts/memory-helper.sh recall "cors" --semantic
# Show auto-capture log
~/.aidevops/agents/scripts/memory-helper.sh log
# View statistics (includes auto-capture counts)
~/.aidevops/agents/scripts/memory-helper.sh stats
# Maintenance
~/.aidevops/agents/scripts/memory-helper.sh validate # Check for stale entries
~/.aidevops/agents/scripts/memory-helper.sh prune # Remove stale memories
# Namespaces (per-runner memory isolation)
~/.aidevops/agents/scripts/memory-helper.sh --namespace my-runner store --content "Runner learning"
~/.aidevops/agents/scripts/memory-helper.sh --namespace my-runner recall "query" --shared
~/.aidevops/agents/scripts/memory-helper.sh namespaces # List all namespaces
Storage: ~/.aidevops/.agent-workspace/memory/memory.db (+ optional embeddings.db for semantic search, namespaces/ for per-runner isolation)
Beyond explicit /remember calls, aidevops can harvest structured session data from each supported AI client's on-disk conversation history. This turns every prior session — regardless of which tool you were using — into searchable context for future sessions.
Per-client storage paths and formats (see runtime-registry.sh):
| Client | Storage path | Format | Default |
|---|---|---|---|
| OpenCode | ~/.local/share/opencode/opencode.db | SQLite | ✅ on |
| Claude Code | ~/.claude/projects/ | JSONL per project | ✅ on |
| Codex CLI | ~/.codex/sessions/ | JSONL, date-partitioned | ✅ on |
| Cursor | ~/Library/Application Support/Cursor/User/workspaceStorage/ | SQLite (state.vscdb) | ✅ on |
| Droid | ~/.factory/sessions/ | JSONL per session | opt-in |
| Gemini CLI | ~/.gemini/tmp/ | JSON per session | opt-in |
| Continue | ~/.continue/sessions/ | JSON per session | opt-in |
| Kilo Code | ~/Library/.../kilocode.kilo-code/tasks/ | JSON (Anthropic schema) | opt-in |
| Kiro | ~/Library/Application Support/Kiro/User/workspaceStorage/ | SQLite | opt-in |
| Kimi CLI | ~/.kimi/sessions/<id>/context.jsonl | JSONL | opt-in |
| Qwen Code | ~/.qwen/tmp/ | JSON per session | opt-in |
| Windsurf | ~/.codeium/windsurf/cascade/ | protobuf (opaque) | ❌ unsupported |
| Amp | server-side (cloud) | needs API auth | ❌ unsupported |
| Aider | <repo>/.aider.chat.history.md | markdown transcript | ❌ unsupported |
Defaults are deliberately conservative. Only the four tier-1 clients — OpenCode, Claude Code, Codex, and Cursor — have memory mining enabled by default. They have the most mature, stable, documented formats and are the ones most likely to contain the full interactive history you'd want to search.
For every other client, memory mining is opt-in per runtime. Enable it with:
AIDEVOPS_FEATURE_MEMORY_GEMINI_CLI=yes setup.sh
AIDEVOPS_FEATURE_MEMORY_CONTINUE=yes setup.sh
Privacy note. Session files contain everything you typed — including secrets, credentials, and file contents pasted into prompts. The mining pipeline runs secretlint-style scrubbing at ingestion and defaults to local-only storage. Never sync the memory DB to a shared or cloud location without reviewing what's in it.
See .agents/memory/README.md for complete documentation.
Slash commands are automatically installed by setup.sh for both OpenCode and Claude Code:
# OpenCode commands deployed to:
~/.config/opencode/command/
# Claude Code commands deployed to:
~/.claude/commands/
# Regenerate commands manually:
.agents/scripts/generate-opencode-commands.sh
.agents/scripts/generate-claude-commands.sh
Both generators read from the same source (.agents/scripts/commands/*.md), ensuring command parity across tools.
In OpenCode or Claude Code, type the command at the prompt:
/preflight
/release minor
/feature add-user-authentication
Commands invoke the corresponding workflow subagent with appropriate context.
User-created agents survive aidevops update. Agents progress through tiers as they mature:
| Tier | Location | Purpose | Survives Update |
|---|---|---|---|
| Draft | ~/.aidevops/agents/draft/ | R&D, experimental, auto-created by orchestration tasks | Yes |
| Custom | ~/.aidevops/agents/custom/ | User's permanent private agents | Yes |
| Shared | .agents/ in repo | Open-source, distributed to all users | Managed by repo |
Promotion workflow: Draft agents that prove useful can be promoted to custom (private) or shared (open-source via PR). Orchestration agents can create drafts in draft/ for reusable parallel processing context.
Create a markdown file in ~/.config/opencode/agent/ (OpenCode) or reference in your AI's system prompt:
---
description: Short description of what this agent does
mode: subagent
temperature: 0.2
tools:
bash: true
specific-mcp_*: true
---
# Agent Name
Detailed instructions for the agent...
See .agents/opencode-integration.md for complete documentation.
# List all servers across providers
./.agents/scripts/servers-helper.sh list
# Connect to specific servers
./.agents/scripts/hostinger-helper.sh connect example.com
./.agents/scripts/hetzner-helper.sh connect main web-server
# Execute commands remotely
./.agents/scripts/hostinger-helper.sh exec example.com "uptime"
# List all monitors
./.agents/scripts/updown-helper.sh list
# Add a new website check
./.agents/scripts/updown-helper.sh add https://example.com "My Website"
# Purchase and configure domain
./.agents/scripts/spaceship-helper.sh purchase example.com
./.agents/scripts/dns-helper.sh cloudflare add-record example.com A 192.168.1.1
# Check domain availability
./.agents/scripts/101domains-helper.sh check-availability example.com
# Basic keyword research with volume, CPC, difficulty
./.agents/scripts/keyword-research-helper.sh research "seo tools" --limit 20
# Google autocomplete long-tail discovery
./.agents/scripts/keyword-research-helper.sh autocomplete "how to" --provider both
# Extended research with SERP weakness detection
./.agents/scripts/keyword-research-helper.sh extended "keywords" --quick
# Competitor keyword research
./.agents/scripts/keyword-research-helper.sh extended --competitor ahrefs.com --limit 50
# Keyword gap analysis (find keywords competitor ranks for but you don't)
./.agents/scripts/keyword-research-helper.sh extended --gap semrush.com,ahrefs.com
# Domain research (all keywords a domain ranks for)
./.agents/scripts/keyword-research-helper.sh extended --domain example.com --limit 100
Features:
# Run quality analysis with auto-fixes
bash .agents/scripts/qlty-cli.sh check 10
bash .agents/scripts/qlty-cli.sh fix
# Run chunked Codacy analysis for large repositories
bash .agents/scripts/codacy-cli-chunked.sh quick # Fast analysis
bash .agents/scripts/codacy-cli-chunked.sh chunked # Full analysis
# AI coding assistance
bash .agents/scripts/ampcode-cli.sh scan ./src
bash .agents/scripts/continue-cli.sh review
# Audit website performance
./.agents/scripts/pagespeed-helper.sh wordpress https://example.com
Wiki Guides:
Agent Guides (in .agents/):
Provider-Specific Guides: Hostinger, Hetzner, Cloudflare, WordPress, Git platforms, Vercel CLI, Coolify CLI, and more in .agents/
aidevops/
├── setup.sh # Main setup script
├── AGENTS.md # AI agent guidance (dev)
├── .agents/ # Agents and documentation
│ ├── AGENTS.md # User guide (deployed to ~/.aidevops/agents/)
│ ├── *.md # 12 primary agents
│ ├── scripts/ # 1,480+ helper scripts
│ ├── tools/ # Cross-domain utilities (video, browser, git, etc.)
│ ├── services/ # External service integrations
│ └── workflows/ # Development process guides
├── configs/ # Configuration templates
├── ssh/ # SSH key management
└── templates/ # Reusable templates
aidevops is configured through two JSONC/JSON files that control framework behaviour, model routing, quality gates, and user preferences. Both are optional -- sensible defaults apply out of the box.
| File | Purpose | CLI |
|---|---|---|
~/.config/aidevops/config.jsonc | Framework config: updates, models, safety, quality, orchestration, paths | aidevops config |
~/.config/aidevops/settings.json | User preferences: onboarding, UI, model routing defaults | settings-helper.sh |
Quick examples:
# Disable automatic updates
aidevops config set updates.auto_update false
# Use thinking-tier verification for destructive operations
aidevops config set safety.verification_tier thinking
# View all current config values
aidevops config list
# Validate config against schema
aidevops config validate
Precedence: Environment variables (AIDEVOPS_*) > config file > built-in defaults. See the full configuration reference for every option, type, default, and example.
Service-specific credentials (hosting, DNS, Git platforms, etc.) use a separate template system:
# 1. Copy and customize configuration templates
cp configs/hostinger-config.json.txt configs/hostinger-config.json
cp configs/hetzner-config.json.txt configs/hetzner-config.json
# Edit with your actual credentials
# 2. Test connections
./.agents/scripts/servers-helper.sh list
# 3. Install MCP integrations (optional)
bash .agents/scripts/setup-mcp-integrations.sh all
See .agents/reference/configuration.md for details on the template system and credential security.
Credential Management:
Manage multiple accounts/clients per service with isolated credential sets:
# Create a new tenant
credential-helper.sh create client-acme
# Switch active tenant
credential-helper.sh switch client-acme
# Set credentials for current tenant
credential-helper.sh set GITHUB_TOKEN ghp_xxx
# Per-project override (gitignored)
echo "client-acme" > .aidevops-tenant
# Export for scripts
source <(credential-helper.sh export)
Resolution priority: Project .aidevops-tenant → Global active tenant → Default
See .agents/tools/credentials/multi-tenant.md for complete documentation.
Quality Assurance:
Contributing:
License: MIT License - see LICENSE file for details Created by Marcus Quinn - Copyright © 2025-2026
For You:
For Your AI Agents:
Get Started:
# npm (recommended)
npm install -g aidevops && aidevops update
# Bun (fast alternative)
bun install -g aidevops && aidevops update
# Homebrew
brew install marcusquinn/tap/aidevops && aidevops update
# Direct from source
bash <(curl -fsSL https://aidevops.sh/install)
An AI operations platform for launching and managing projects across every business domain — from code to content, infrastructure to invoicing.
FAQs
AI DevOps Framework - AI-assisted development workflows, code quality, and deployment automation
The npm package aidevops receives a total of 6,778 weekly downloads. As such, aidevops popularity was classified as popular.
We found that aidevops demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Research
/Security News
Benign-looking npm packages split malicious functionality across a dependency chain that deploys a cross-platform RAT targeting Alibaba developers.
Research
/Security News
Two Joyfill npm beta releases contain an import-time implant that uses blockchain transactions to retrieve a remote-access trojan.
Security News
Socket releases free Certified Patches for high-severity Nuxt vulnerabilities, including server-side remote code execution through server island props.