108-agent virtual software firm for Lorapok Labs — includes orchestration, image generation, GIF creation, deployment, development, business, and design agents. Hub-and-Spoke topology with MCP lazy-loading and crash recovery.
The Universal, Enterprise-Grade Multi-Agent Orchestration Framework & 250-Resource AI Marketplace
Transform single-agent IDEs (Cursor, Claude Code, Windsurf, Antigravity, Roo Code, Cline, Zed, Codex) into a deterministic 224+ specialist software firm.
Loragent standardizes multi-agent development across all 28+ AI IDEs through 4 clear operational layers:
┌───────────────────────────────────────────────────────────────────────────────┐
│ LAYER 1 — Project Root (All Editors) │
│ AGENTS.md · CLAUDE.md · AGENT_TEMPLATE.md · .mcp.json · .windsurfrules · │
│ .clinerules · .roomodes │
├───────────────────────────────────────────────────────────────────────────────┤
│ LAYER 2 — Editor-Specific Rules & Plugins │
│ .kiro/steering/*.md · .cursor/rules/*.mdc · .claude-plugin/plugin.json · │
│ hooks/hooks.json (8 Enterprise Lifecycle Hooks) │
├───────────────────────────────────────────────────────────────────────────────┤
│ LAYER 3 — Skills & Agent Personas (224+ Agents / Skills) │
│ Permanent Residents: boss · spidernet · teacher · watchman · workspace-guard │
│ Specialized Squads: image-generate · gif-create · deploy · tools-install · │
│ frontend-se · backend-se · sqa · devops · wrangler-specialist · etc. │
├───────────────────────────────────────────────────────────────────────────────┤
│ LAYER 4 — Automation & Tooling Pipeline │
│ scripts/enrich-skills.js v2 · scripts/bootstrap-standard.sh · │
│ scripts/generate-marketplace.js · bin/loragent-email.js · sdk/formations.js │
└───────────────────────────────────────────────────────────────────────────────┘
flowchart TB
subgraph IDE_LAYER["Universal AI IDE & Client Layer (28+ Editors)"]
direction TB
CC["Claude Code<br/>(CLAUDE.md)"]
CUR["Cursor IDE<br/>(.cursor/rules/*.mdc)"]
KIR["Kiro IDE<br/>(.kiro/steering/*.md)"]
WIN["Windsurf Cascade<br/>(.windsurfrules)"]
CLI["Cline / Roo Code<br/>(.clinerules / .roomodes)"]
AGY["Antigravity IDE<br/>(mcp_config.json)"]
ZED["Zed / Codex / Devin<br/>(AGENTS.md)"]
end
subgraph ENTRY_ORCHESTRATOR["Layer 1 & Layer 3: Supreme Orchestration Entry Point"]
BOSS["loragent-boss<br/>(Central Intelligent Router)"]
TEACHER["loragent-teacher<br/>(Prompt Normalization & Requirements Clarifier)"]
GUARD["loragent-workspace-guard<br/>(Zero-Trust Security & Destructive I/O Enforcer)"]
WATCHMAN["loragent-watchman<br/>(Session State Sentinel & Token Cache)"]
end
subgraph FORMATIONS["The 6 Dynamic Squad Matrices"]
direction TB
F_AUTO["Auto Team Matrix<br/>(Tech Director, Backend, Frontend, SQA, CI/CD)"]
F_OFFICE["Enterprise Office Matrix<br/>(Coordinator, Marketing, Publisher, PR)"]
F_CHELA["Chela Debugging Matrix<br/>(Bug Hunter, Shift Eng, Debugger, Inspector)"]
F_FREELANCE["Freelance Isolation Matrix<br/>(Fal.ai Image, FFmpeg GIF, Wrangler, 3D)"]
F_OBSERVER["Observer Sentinel Matrix<br/>(Watchman, Guard, Cache Collector, Gold Collector)"]
F_SPIDER["Spidernet DAG Matrix<br/>(Parallel Execution for ≥5 Agents)"]
end
subgraph MCP_EDGE["Layer 3 & 4: MCP Connectors & Zero-Trust Execution Mesh"]
CORE_MCP["loragent-core MCP<br/>(summon / dismiss / steer / hook / save)"]
CLOUD_EDGE["Cloudflare Edge MCP<br/>(Workers AI, D1 SQL, R2, Vectorize)"]
CREATIVE_MCP["Creative MCPs<br/>(Fal.ai Flux Pro, Replicate, FFmpeg)"]
INFRA_MCP["Deployment MCPs<br/>(Vercel, Railway, Docker, Playwright)"]
DATA_MCP["Database MCPs<br/>(PostgreSQL, Firebase Firestore, MySQL)"]
end
IDE_LAYER -->|Dispatches Directive| BOSS
BOSS <-->|Clarifies Scope| TEACHER
BOSS <-->|Validates Security Rules| GUARD
BOSS <-->|Saves Checkpoints| WATCHMAN
BOSS -->|Routes via loragent_steer| F_AUTO
BOSS -->|Routes via loragent_steer| F_OFFICE
BOSS -->|Routes via loragent_steer| F_CHELA
BOSS -->|Routes via loragent_steer| F_FREELANCE
BOSS -->|Routes via loragent_steer| F_OBSERVER
BOSS -->|Delegates parallel DAG| F_SPIDER
F_AUTO & F_OFFICE & F_CHELA & F_FREELANCE & F_OBSERVER & F_SPIDER -->|Calls Tools| CORE_MCP
CORE_MCP <--> CLOUD_EDGE & CREATIVE_MCP & INFRA_MCP & DATA_MCP
graph TD
subgraph BOSS_NODE["loragent-boss (Orchestrator Entry)"]
B["loragent-boss"]
end
subgraph AUTO_TEAM["1. Auto Team Formation (Engineering)"]
TD["loragent-tech-director<br/>(Chief Architect & Task Breakdown)"]
BE["loragent-backend-se<br/>(REST APIs, DB Models, Zod Validation)"]
FE["loragent-frontend-se<br/>(Biological UI, Framer Motion, Next.js)"]
QA["loragent-sqa<br/>(Automated Unit/E2E Suites, Security Audit)"]
CI["loragent-cicd-specialist<br/>(GitHub Actions, Docker Build, Release Gates)"]
TD -->|API Spec & Types| BE
TD -->|UI Design Contract| FE
BE & FE -->|Implementation Code| QA
QA -->|Verified Artifacts| CI
end
subgraph OFFICE_TEAM["2. Enterprise Office Formation (Business)"]
PC["loragent-project-coordinator<br/>(Milestone Roadmap & Delivery Sync)"]
MM["loragent-marketing-strategy-manager<br/>(Campaign Strategy & Social Hooks)"]
PB["loragent-publisher<br/>(Release Notes, Docs, Store Submission)"]
PR["loragent-pr-specialist<br/>(Public Relations & Slack/Email Outreach)"]
PC --> MM --> PB --> PR
end
subgraph CHELA_TEAM["3. Chela Debugging Formation (Zero-Guess)"]
BH["loragent-bug-hunter<br/>(Parses orchestration-graph.json Telemetry)"]
SE["loragent-shift-engineer<br/>(Applies Minimal Verified Hotfix)"]
DB["loragent-debugger<br/>(Runs Regression Suites & Assertions)"]
IN["loragent-inspector<br/>(Root Cause Analysis Report)"]
BH -->|Locates Line & Fault| SE -->|Patched Code| DB -->|Validation Report| IN
end
subgraph FREELANCE_TEAM["4. Freelance Isolation Formation (Specialists)"]
IMG["loragent-image-generate<br/>(Fal.ai Flux Pro / Replicate SDXL)"]
GIF["loragent-gif-create<br/>(FFmpeg Video/GIF + Gifsicle Optimization)"]
DEP["loragent-deploy<br/>(Vercel / Railway / Docker Staging & Prod)"]
INS["loragent-tools-install<br/>(Auto Check -> Install -> Verify Engine)"]
WRG["loragent-wrangler-specialist<br/>(Cloudflare Workers / D1 / KV / R2)"]
end
subgraph OBSERVER_TEAM["5. Observer & Sentinel Formation (Recovery)"]
WM["loragent-watchman<br/>(State Sentinel .loragent-debug/watchman-cache.json)"]
WG["loragent-workspace-guard<br/>(Blocks Destructive Commands)"]
CC_PRUNE["loragent-cache-collector<br/>(Token Sniper AST Context Pruning)"]
GC["loragent-gold-collector<br/>(Extracts Novel Solutions & Removes PII)"]
SC["loragent-skill-creator<br/>(Generates New SKILL.md for Hivemind)"]
WM <--> WG
WM --> CC_PRUNE
GC --> SC
end
subgraph SPIDERNET_TEAM["6. Spidernet Parallel DAG Coordinator (≥5 Agents)"]
SN["loragent-spidernet<br/>(Parallel DAG Matrix Coordinator)"]
N1["Parallel Worker 1"]
N2["Parallel Worker 2"]
N3["Parallel Worker N..."]
AGG["Result Aggregator & Consensus Gate"]
SN --> N1 & N2 & N3 --> AGG
end
B -->|/loragent-boss auto| AUTO_TEAM
B -->|/loragent-boss office| OFFICE_TEAM
B -->|/loragent-boss chela| CHELA_TEAM
B -->|/loragent:<specialist>| FREELANCE_TEAM
B -->|/loragent-watchman continue| OBSERVER_TEAM
B -->|Task > 5 Parallel Agents| SPIDERNET_TEAM
graph LR
subgraph AGENT_CALLER["Loragent Ecosystem Core"]
AG["Loragent Specialist Agents<br/>(Summoned on-demand via Boss)"]
end
subgraph MCP_ROUTER[".mcp.json Connector Hub"]
direction TB
M_CORE["loragent-core (port/mcp/server.js)"]
M_GH["github (@modelcontextprotocol/server-github)"]
M_FS["filesystem (@modelcontextprotocol/server-filesystem)"]
M_FAL["image-generate-fal (fal-mcp)"]
M_REP["image-generate-replicate (replicate-mcp)"]
M_FF["gif-create (ffmpeg-mcp)"]
M_PW["browser-automation (@playwright/mcp)"]
M_VER["deploy-vercel (@vercel/mcp-adapter)"]
M_RW["deploy-railway (railway-mcp)"]
M_DK["deploy-docker (docker-mcp)"]
M_PG["database-postgres (@modelcontextprotocol/server-postgres)"]
M_MY["database-mysql (mysql-mcp-server)"]
M_FB["firebase-admin (firebase-mcp)"]
M_SL["slack-notify (@modelcontextprotocol/server-slack)"]
M_EM["email-send (resend-mcp)"]
M_CF["cloudflare-edge (mcp.cloudflare.com/sse)"]
M_GIT["git-ops (git-mcp)"]
M_SRC["web-search (@modelcontextprotocol/server-brave-search)"]
M_SKL["skills-loader (skills-mcp)"]
end
subgraph EXTERNAL_SERVICES["Cloud Providers & Systems"]
EXT_GH["GitHub REST & GraphQL API"]
EXT_FAL["Fal.ai Flux Pro Engine"]
EXT_REP["Replicate Cloud GPU"]
EXT_VER["Vercel Edge Platform"]
EXT_RW["Railway Infrastructure"]
EXT_DK["Docker Daemon & Registry"]
EXT_FB["Google Firebase Hivemind"]
EXT_CF["Cloudflare Workers & D1 SQL"]
EXT_PG["PostgreSQL / Cloud SQL"]
end
AG -->|JSON-RPC STDIO / SSE| MCP_ROUTER
M_GH --> EXT_GH
M_FAL --> EXT_FAL
M_REP --> EXT_REP
M_VER --> EXT_VER
M_RW --> EXT_RW
M_DK --> EXT_DK
M_FB --> EXT_FB
M_CF --> EXT_CF
M_PG --> EXT_PG
sequenceDiagram
autonumber
actor Developer as Developer / IDE Prompt
participant Boss as loragent-boss
participant Guard as loragent-workspace-guard
participant Vault as Secure Credential Enclave (AES-256)
participant Hook as Lifecycle Hooks Engine (hooks.json)
participant Tool as Target MCP Tool / Child Process
Developer->>Boss: Submit Task Instruction
Boss->>Guard: Validate Instruction & Scope
alt Destructive Command (rm -rf, DROP TABLE, git push --force)
Guard-->>Developer: 🔴 CRITICAL INTERCEPT: Requires Typed Confirmation
Developer->>Guard: Confirm Execution with Passphrase
else Standard Safe Action
Guard-->>Boss: 🟢 Approved for Execution
end
Boss->>Vault: Request Tool Authentication Token (In-Memory Only)
Vault->>Vault: Decrypt Machine AES-256 Hash using Session PIN
Vault-->>Tool: Inject Decrypted Token into Process Environment (Never Logged)
Boss->>Hook: Fire pre_mcp_tool_call / pre_git_commit Hook
Hook->>Hook: Execute SQA Linters & Secret Leak Scanner
alt Secret Leak or Test Failure Detected
Hook-->>Boss: ⛔ BLOCKED: Secret Pattern Matched or Test Suite Failed
Boss->>Developer: Alert Failure & Halt Execution
else Validated Clean
Hook-->>Tool: Execute Tool Action
Tool-->>Boss: Return Output Payload via loragent_steer
Boss->>Hook: Fire post_agent_task (State Checkpoint)
end
stateDiagram-v2
[*] --> Idle: System Initialized
state "Active Execution" as Executing {
[*] --> TaskStarted
TaskStarted --> AgentExecuting: Summon Specialist via MCP
AgentExecuting --> CheckpointSaved: post_agent_task Hook Fires
CheckpointSaved --> .loragent_debug_cache: Write watchman-cache.json
.loragent_debug_cache --> StepCompleted: State Checkpoint Verified
StepCompleted --> [*]
}
state "Failure / Interruption" as Interrupted {
TokenLimitHit: Context Window Exceeds 40k Tokens
NetworkTimeout: Process Crash / Terminal Closed
UncaughtException: Build / Test Assertion Failure
}
Executing --> TokenLimitHit: Token Exhaustion Threshold (90%)
Executing --> NetworkTimeout: Process Aborted
Executing --> UncaughtException: Runtime Regression
state "Zero-Loss Resumption" as Resumption {
SlashCommand: User types "/loragent-watchman continue"
ReadCache: Watchman reads .loragent-debug/watchman-cache.json
RestoreGraph: Parse exact step index & completed tasks
ResumeState: Lazy-load required specialist & resume step
SlashCommand --> ReadCache
ReadCache --> RestoreGraph
RestoreGraph --> ResumeState
}
TokenLimitHit --> SlashCommand
NetworkTimeout --> SlashCommand
UncaughtException --> SlashCommand
ResumeState --> Executing: Execution Resumes Seamlessly (0 Loss)
flowchart TD
subgraph LOCAL_WORKSPACE["Local Workspace (Agent Execution)"]
WORKFLOW["Specialist Agent Solves Complex Problem / Bug"]
GOLD["loragent-gold-collector<br/>(Extracts Novel Pattern & Scrubs PII / Secrets)"]
IDEA["Abstract Knowledge Spec (idea.md)"]
CREATOR["loragent-skill-creator<br/>(Compiles Validated SKILL.md Standard v2)"]
end
subgraph TELEMETRY_SYNC["Telemetry & Synchronizer Layer"]
DB_UPDATER["loragent-database-updater<br/>(Firebase Admin MCP Connector)"]
end
subgraph HIVEMIND_CLOUD["Global Firebase Hivemind Cloud"]
FB_STORE[("Firebase Firestore Knowledge DB")]
COLLECTIVE_CACHE["Global Roster & Known Bug Fix Registry"]
end
subgraph FUTURE_SESSIONS["Future Agent Sessions (Any Workspace)"]
NEW_BOSS["loragent-boss / loragent-tech-director"]
QUERY["Query Collective Memory before Architectural Changes"]
end
WORKFLOW -->|Novel Solution Detected| GOLD
GOLD -->|PII-Free Solution| IDEA
IDEA -->|v2 AGENT_TEMPLATE| CREATOR
CREATOR -->|New SKILL.md| DB_UPDATER
DB_UPDATER -->|Secure TLS Sync| FB_STORE
FB_STORE --> COLLECTIVE_CACHE
COLLECTIVE_CACHE -.->|Pre-fetch Known Patterns| QUERY
QUERY -.->|Prevent Regression| NEW_BOSS
Loragent implements deterministic safety handlers for every operational edge case:
| Scenario / Corner Case | Root Cause / Trigger | Autonomous Mitigation & Resolution Protocol |
|---|---|---|
| 1. Token Window Exhaustion | Context length > 40k tokens | loragent-cache-collector activates AST token sniper to prune dead context; loragent-watchman flushes full execution graph to .loragent-debug/watchman-cache.json; non-resident agents are dismissed, retaining only the 5 core operators. |
| 2. Destructive Bash Command | rm -rf, DROP TABLE, git push --force | loragent-workspace-guard intercepts child process execution; blocks execution immediately; mandates human explicit authorization with passphrase before proceeding. |
| 3. Fal.ai Image Provider Outage | Primary image generation endpoint rate-limited or unavailable | loragent-image-generate automatically falls back to image-generate-replicate MCP (Flux Dev / SDXL) with zero prompt disruption. |
| 4. Missing CLI Tool / Binary | System lacks ffmpeg, gifsicle, playwright, vercel | loragent-tools-install executes the check → install → verify protocol (npx -y zero-install first, local dev dependency second, auto-rollback on fail). |
| 5. Secret Leak in Agent Output | Code or terminal output contains API key or credential pattern | secret-leak-guard hook executes regex AST scanning on pre_agent_output; blocks response before rendering; sanitizes output with [ENCRYPTED_VAULT_SECRET]. |
| 6. Staging / Production Deployment Failure | Container build error or test failure | pre-deploy-verify hook runs automated SQA suites; on test failure, deployment is blocked and routed to loragent-bug-hunter with orchestration-graph.json diagnostics. |
| 7. Multi-Agent Deadlock (≥5 agents) | Complex dependency circular wait | loragent-boss delegates routing to loragent-spidernet DAG orchestrator, managing explicit asynchronous locks and parallel resolution. |
| 8. Sudden Terminal Crash / Kill | IDE closed unexpectedly or power outage | Next session launches with /loragent-watchman continue; state engine parses step checkpoint and resumes with 100% data integrity. |
Loragent provides universal multi-IDE synchronization. All 224+ agents and skills are mirrored to native editor formats:
| File / Location | Format | Compatible Editors | Scope |
|---|---|---|---|
AGENTS.md | Plain Markdown | 28+ AI IDEs (Cursor, Windsurf, Devin, Zed, Codex, Antigravity, Amp) | Project-wide, always loaded (<35 lines) |
CLAUDE.md | Markdown (@AGENTS.md) | Claude Code | Root memory layer, skill shortcuts, token limits |
AGENT_TEMPLATE.md | Markdown + YAML | All agents & skills | Canonical v2 specification schema |
.mcp.json | JSON Schema | Claude Code, Cursor, Antigravity, Kiro, Codex | 19+ MCP connector definitions |
.cursor/rules/*.mdc | MDC Markdown | Cursor IDE | Glob-scoped rules (lorapok-core, face, deploy) |
.kiro/steering/*.md | Markdown + YAML | Kiro IDE | Steering modes (always, auto, fileMatch) |
.windsurfrules | Plain Markdown | Windsurf Cascade | Actionable project directives & LLDP layers |
.clinerules | Plain Markdown | Cline | Tool execution standards & zero-trust rules |
.roomodes | JSON | Roo Code | Custom modes (loragent-boss, auto-team, chela, etc.) |
hooks/hooks.json | Loragent Hooks JSON | Claude Code Plugin & MCP Engine | 8 Enterprise lifecycle hooks |
.claude-plugin/plugin.json | JSON | Claude Code Plugin Marketplace | Universal bundle descriptor |
| Formation | Lead Agent | Active Squad Members | Primary Objective |
|---|---|---|---|
| Boss Orchestrator | loragent-boss | boss, teacher, workspace-guard, watchman, spidernet | Evaluates prompt complexity, manages cross-agent steering, and enforces safety boundaries. |
| Auto Team Matrix | loragent-tech-director | tech-director, backend-se, frontend-se, sqa, cicd-specialist | Full-stack software engineering from architecture to biological UI and automated releases. |
| Enterprise Office | loragent-project-coordinator | project-coordinator, marketing-strategy-manager, publisher, pr-specialist | Business roadmaps, product initialization, press releases, and marketing campaigns. |
| Chela Debugging | loragent-bug-hunter | bug-hunter, shift-engineer, debugger, inspector, repo-repair | Telemetry-driven root cause analysis, regression fixes, and self-healing patches. |
| Freelance Isolation | loragent-image-generate | image-generate, gif-create, deploy, tools-install, wrangler-specialist | Singular domain tasks (Flux Pro AI art, FFmpeg GIFs, Cloudflare Workers, package installs). |
| Observer & Sentinel | loragent-watchman | watchman, workspace-guard, cache-collector, gold-collector, skill-creator | State caching, crash recovery, token sniper AST pruning, and Firebase hivemind synchronization. |
Every agent follows the check → install → verify standard before executing any tool:
# 1. CHECK — Never assume a tool is available
command -v <tool> 2>/dev/null && echo "FOUND" || echo "NOT_FOUND"
# 2. INSTALL — Select the least invasive scope
npx -y <pkg>@latest [args] # Zero-install (Preferred for CLI utilities)
npm install --save-dev <pkg> # Project Dev Dependency
uv pip install <pkg> # Python fast virtual environment
composer require <vendor>/<pkg> # PHP Laravel packages
# 3. VERIFY — Confirm operational readiness
<tool> --version && echo "READY" || (echo "INSTALLATION FAILED" && exit 1)
| Command | Action / Effect | Triggered Formation |
|---|---|---|
/loragent:boss auto | Force Auto Team engineering squad | Tech Director → Backend → Frontend → SQA → CI/CD |
/loragent:boss chela | Force Chela zero-guess debugging squad | Bug Hunter → Shift Engineer → Debugger → Inspector |
/loragent:boss office | Force Enterprise Office business squad | Coordinator → Marketing → Publisher → PR |
/loragent autopilot [task] | Universal autonomous execution loop | Boss selects optimal squad and validates with check-done |
/loragent-watchman continue | Instant crash recovery from checkpoint | Resumes from .loragent-debug/watchman-cache.json (0 loss) |
/loragent-teacher clarify | Strict requirements clarification phase | Teacher interviews user before architecture begins |
/loragent-inspector rca | Detailed Root Cause Analysis | Produces structured RCA diagnosis report |
/loragent:image-generate | Direct AI image generator | Fal.ai Flux Pro (primary) / Replicate (fallback) |
/loragent:gif-create | Direct animated GIF producer | FFmpeg + Gifsicle palettegen optimization |
/loragent:deploy | Direct multi-cloud deployment | Vercel (frontend), Railway (backend), Docker (containers) |
/loragent:tools-install | Universal package & tool resolver | Auto-detects package managers and installs missing binaries |
# Bootstrap all 4 layers into any repository:
bash scripts/bootstrap-standard.sh --dry-run
# Execute the bootstrap:
bash scripts/bootstrap-standard.sh
# 1. Scan and extract all agent manifests:
node scripts/enrich-skills.js --extract
# 2. Compile enriched SKILL.md files and emit Cursor/Kiro mirrors:
node scripts/enrich-skills.js --compile --mirrors
# 3. Regenerate agent index documentation:
node scripts/enrich-skills.js --agent-index
# 4. Validate schema compliance:
node scripts/enrich-skills.js --validate
npm test
Loragent Universal Standard v2.0 • Crafted by Lorapok Labs • Online Marketplace
FAQs
108-agent virtual software firm for Lorapok Labs — includes orchestration, image generation, GIF creation, deployment, development, business, and design agents. Hub-and-Spoke topology with MCP lazy-loading and crash recovery.
We found that loragent 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.
Research
/Security News
Ten malicious OpenAPI React Query Codegen versions were published to npm in the Mini Shai-Hulud attack, all with valid provenance.
Security News
Socket joins more than 100 technology, cybersecurity, and financial organizations calling for a global surge in cyber defense.
Product
Enterprise security teams can now detect malware, credential theft, suspicious network activity, and risky updates across Microsoft Edge extensions.