@bobsworkshop/cli
Advanced tools
Bob's CLI is a locally-installed developer tool that provides a senior-level AI engineering partner directly inside your native terminal. Stay in your development environment. Never switch to a browser. Ship faster.
Installation · Quick Start · Features · The Crew · VaultBob · UserBob · Command Center · Cross-Project References · Docs
Built by Bob's Workshop — A Seedling Company
Every other AI coding assistant lives in a browser, disconnected from your actual workflow. Bob lives where your code lives — in your terminal. He sees your files, understands your architecture, writes code with your approval, and learns how YOU work over time.
We believe AI should empower developers, not exploit them. Your code stays on your machine. Your data belongs to you. No surveillance, no vendor lock-in, no token meters draining your wallet while you think. Bob's CLI is built for the developer who refuses to hand over sovereignty in exchange for convenience. This is AI that works FOR you — on YOUR hardware, under YOUR control, at YOUR pace. Power to the people who build.
| Feature | Bob's CLI | Claude Code | Copilot | Cursor |
|---|---|---|---|---|
| Local file awareness | ✅ | ✅ | ✅ | ✅ |
| Zero-cost local model | ✅ | ✅ | ❌ | ❌ |
| Behavioral profiling | ✅ | ❌ | ❌ | ❌ |
| Personalization Mode | ✅ | ❌ | ❌ | ❌ |
| Digital twin simulation | ✅ | ❌ | ❌ | ❌ |
| Autonomous task dispatch | ✅ | ❌ | ❌ | ❌ |
| Multi-agent orchestration | ✅ | ❌ | ❌ | ❌ |
| Supervised commit review | ✅ | ❌ | ❌ | ❌ |
| Conversation persistence | ✅ | ✅ | ❌ | Partial |
| Deep Dives & Forks | ✅ | ❌ | ❌ | ❌ |
| Remote execution (SovereignLink) | ✅ | Partial | ❌ | ❌ |
| Cross-surface sync (CLI ↔ Web) | ✅ | ✅ | ❌ | ❌ |
| Autonomous code repair | ✅ | ✅ | ❌ | ✅ |
| Source code stays on-device | ✅ | ✅ | ❌ | ✅ |
| Encrypted cloud backup (VaultBob) | ✅ | ❌ | ❌ | ❌ |
| Per-file surgical restore | ✅ | ❌ | ❌ | ❌ |
| Full machine migration | ✅ | ❌ | ❌ | ❌ |
| Cross-project references | ✅ | ❌ | ❌ | ❌ |
pnpm add -g @bobsworkshop/cli
Or with npm:
npm install -g @bobsworkshop/cli
Verify:
bob whoami
Requirements:
📖 Full setup guide: https://seedling-io.gitbook.io/bob-cli/
bob chat "hello, what can you help me with?"
Bob auto-detects Ollama running on your machine. No configuration needed. No internet. No API keys. No cost. Your code never leaves your machine.
bob login
bob chat "help me refactor this service"
Sync to web. Access Claude, Gemini, deep dives, forks, personalization, and UserBob.
When you first install Bob's CLI, you're greeted with a branded welcome screen:
| Feature | Description |
|---|---|
| Chat | AI coding partner with automatic file discovery |
| Consult | Strategic advice, no code output |
| Index | AI-powered project understanding |
| Analyse | Full QA code review with auto-fix |
| Autonomy | Autonomous repair across entire codebase |
| Profile | Behavioral DNA profiling + dashboard |
| The Crew | Multi-agent orchestration — spawn a team, set a mission, ship code |
| VaultBob | Encrypted backup, versioning & restore — your code's permanent memory |
| UserBob | AI digital twin simulation — your autonomous proxy |
| Command Center | Inspect, approve, and manage autonomous task dispatch |
| Deep Dive | Sandboxed exploration on any message |
| Fork | Branch conversations into sub-projects |
| SovereignLink | Remote execution from any device |
| BYOK | Bring your own API keys |
| Push | Git stage + commit + push in one command |
| Cross-Project References | Pull in context from other org projects mid-conversation |
Bob performs production-grade QA reviews across your entire codebase — identifying bugs, features, improvements, and upgrades with actionable implementation instructions:
bob analyse # Run full code review
bob analyse --results # View dashboard
bob analyse --auto # Auto-fix with safety constraints
v1.0.0 introduces The Crew — a fully local autonomous multi-agent orchestration system that runs entirely on your hardware. Spawn specialized agents, set a mission, and let DirectorBob coordinate the team. The Crew plans, implements, reviews, and commits real code changes to your project — with you in control the entire time.
bob agent-run "Add authentication to the app"
DirectorBob reads your actual codebase, decomposes the mission into specific tasks, and dispatches agents to execute in parallel or sequence based on dependencies. Every file an agent writes is reviewed by DirectorBob before the task is marked complete. Every commit is reviewed before it touches git. Every file is backed up before it is modified.
╔══════════════════════════════════════════════════════════╗
║ 🎬 DirectorBob — Autonomous Mission Control
╠══════════════════════════════════════════════════════════╣
║ Add authentication to the app
║ Mission: m_1782095907390 │ Tasks: 3 │ Team: 2 agents
║
║ @builderBob @architectBob
╠══════════════════════════════════════════════════════════╣
║ 📋 Task Map — 3 tasks
║ ⏸ @architectBob CREATE lib/auth/auth_service.dart
║ ⏸ @builderBob PATCH lib/main.dart: ADD auth init ← 1 dep
║ ⏸ @builderBob PATCH lib/screens/login.dart ← 1 dep
╚══════════════════════════════════════════════════════════╝
[9:14:02 AM] 🎬 DirectorBob: Dispatching @architectBob → [CREATE]
[9:14:02 AM] 🎬 DirectorBob: Dispatching @builderBob → [PATCH]
@architectBob Created lib/auth/auth_service.dart (42 lines)
✅ DirectorBob approved @architectBob's work
✅ DirectorBob approved ✅ committed: 91a039d
@builderBob Patched lib/main.dart — auth init added
✅ DirectorBob approved @builderBob's work
DirectorBob is your Head of Engineering. Before dispatching a single task he reads your project's file tree, assesses what exists and what doesn't, and generates a dependency-aware task map. He decides who does what, in what order, and in what way — CREATE for new files, PATCH for surgical edits, REFACTOR for structural changes, REPLACE for placeholder files.
Agents execute their assigned tasks using the same file-writing pipeline that powers bob chat — proven, approval-gated, and backed up on every write. They self-evaluate their own work with a satisfaction score. If the score isn't high enough, they retry. If they stagnate, DirectorBob intervenes with specific feedback. If they still can't resolve it, the mission pauses and asks you.
Every file write is reviewed by DirectorBob before the task is marked complete. He reads the original file, the new file, the project context, and the task instruction. If an agent patched 5 lines when the task required 300, or gutted a working file instead of extending it — denied, restored from backup, retried with feedback.
Every commit is reviewed before it touches git. When an agent calls gitCommit, DirectorBob performs a full code review against the task instruction. APPROVE and the commit executes. DENY and the original is restored, a revision note is injected, and the agent retries. Three denials and the mission pauses for you.
bob agent spawn builder "Implement features and write code"
bob agent spawn architect "Design approach and review structure"
bob agent list # See your active team
bob agent status # Detailed status with session info
bob agent hub # Interactive command center
bob agent-run "your mission" # Launch with active team
bob agent-run --dry-run "your mission" # Preview task map only
bob agent-run --resume # Resume a paused mission
bob agent-run --no-commit # Skip post-mission commit prompt
/pause — pause after active tasks complete
/resume — resume from pause
/status — full task map with current state
/skip <taskId> — skip a stuck task
/inject "note" — send a director note mid-mission
/set-target <agent> <n> — adjust satisfaction target
/abort — stop everything immediately
Agents can be given specialized personas that shape how they think and communicate:
bob agent spawn architect "Design the system" --persona local:architectBob
bob agent spawn security "Audit the codebase" --persona local:securityBob
bob agent personas # List all available personas
| Persona | Specialty |
|---|---|
local:architectBob | Contract-first, systems design |
local:builderBob | Ships fast, pragmatic execution |
local:qaEngineerBob | Testing, edge cases, reliability |
local:securityBob | Threat modeling, zero trust |
local:frontendBob | UI/UX, accessibility, components |
local:backendBob | APIs, reliability, idempotency |
local:devopsBob | CI/CD, infrastructure, pipelines |
Every other agent system executes and hopes for the best. The Crew has DirectorBob — a supervisor who reviews every file change before it's accepted, every commit before it's recorded, and every stuck task before it spirals. Your codebase is protected at every step. Your backups are automatic. Your control is never surrendered.
Runs entirely on your local model. Zero cloud required. Zero cost.
v0.7.0 introduces VaultBob — the only backup, versioning, and restore system built specifically for how developers work. Every conversation you've had with Bob, every project Bob has learned, every version of every file — encrypted, safe, and accessible from anywhere in the world.
Built in partnership with AWS — the same security infrastructure trusted by the world's largest financial institutions, government agencies, and the companies building the future. Your data is encrypted on your machine before it ever leaves. AWS never sees your code. Nobody does except you.
bob backup create # Back up Bob's knowledge of your project
bob backup create --source # Back up your actual code files
bob backup create --source --file <path> # Back up one specific file (surgical)
bob backup create --global # Back up everything — all projects (Grid)
bob backup create --archive "before-auth-refactor"
bob backup create --source --archive "v1.0-launch-state"
Think of named archives like Git tags. Regular backups are your auto-save. Named archives are your intentional saves — findable by name, not just by date.
bob backup list # See all revisions for this project
bob backup list --source # See all source code snapshots
bob backup list --source --file <path> # See one file's complete history
bob backup restore # Interactive picker — choose any version
bob backup restore --source # Restore source code
bob backup restore --source --file <path> # Restore one specific file
bob backup restore --global # Full machine restore (Grid)
Before every restore, VaultBob automatically saves your current state locally. You cannot permanently lose your work through normal use of VaultBob.
bob backup create --global
Back up your entire engineering brain — every project, every conversation, every behavioral profile, every analysis. When you get a new laptop, run bob backup restore --global and you're back. Everything. In minutes. Not days.
"New laptop. Same Bob. Same context. Same you."
| Workshop Plan | Archive Slots | Retention Period |
|---|---|---|
| Patch | 0 slots | 1 month |
| Build | 3 slots | 3 months |
| Forge | 6 slots | 6 months |
| Grid | 12 slots | 12 months |
VaultBob integrates directly with SovereignLink — back up and restore your remote machine without being physically present:
bob remote backup # Trigger backup on your home machine
bob remote backup --source # Remote source backup
bob remote restore # Restore latest on remote machine
v0.6.0 introduces UserBob — the most advanced feature in Bob's CLI. UserBob creates an autonomous AI proxy of you, built from your behavioral DNA, engineering philosophy, and communication style. Your digital twin negotiates with Bob on your behalf to advance a mission you define. You watch. You tune. You approve the results.
bob userbob "Refactor the auth service error handling"
The simulation runs autonomously — no human input required. Bob and your digital twin negotiate until satisfaction reaches your target, then implementation tasks are dispatched to Mini Bob for execution.
┌─ UserBob ──────────────────────────────────────────────────┐
│ The error handling in AuthService is incomplete. Bob, │
│ show me the current implementation before we proceed. │
└────────────────────────────────────────────────────────────┘
[SAT: 42%] [RES: 78%] [CONVERGING]
┌──────────────────────────── Bob ─┐
│ Here's the current auth service │
│ — I can see three areas where │
│ error handling is missing... │
└──────────────────────────────────┘
─── MISSION CONTROL ──────────────────────────────────────────
SAT: 42% → 85% │ STAG: 0/3 │ DIV: 0/2 │ GRADE: 60
────────────────────────────────────────────────────────────────
Options:
bob userbob "mission" # Inline mission
bob userbob # Interactive mission prompt
bob userbob --target 70 --grading 60 # Custom parameters
bob userbob --stag 3 --div 2 # Set safety thresholds
bob userbob --resume # Resume stalled session
bob userbob --local "mission" # Tier 1 Ollama mode
Mid-session slash commands:
/set target 80 Update satisfaction target
/set grading 70 Update Teacher's Curve
/set stag 5 Set stalemate threshold
/set div 3 Set divergence threshold
/inject "note" Steer the simulation mid-session
/status Show current parameters
/abort Stop immediately
Generate your behavioral DNA first for best results:
bob profile --today # Local (requires Ollama)
bob profile --cloud # Cloud (Power tier)
Every task UserBob dispatches to Mini Bob is visible, manageable, and auditable from the CLI:
bob command-center # Interactive task board
bob cc # Alias
bob command-center --stream # Live decision stream
bob command-center --settings # Configure autonomy thresholds
What you see:
─── COMMAND CENTER ──────────────────────────────────────────
2 PENDING │ 8 RUNNING │ 31 DONE │ 41 TOTAL
────────────────────────────────────────────────────────────────
● NEEDS APPROVAL [frontend ] Create the TabletHomePage layout...
● IN PROGRESS [backend ] Update auth service error handling...
● COMPLETE [cloud_functions] Deploy rate limiter utility...
Select any task to see the full chain of custody: Trigger → Request → Outcome. Approve or deny pending tasks directly from the terminal with live execution log streaming.
Configure how much autonomy UserBob has:
bob command-center --settings
# Set global confidence threshold (tasks below this % require approval)
# Set per-category overrides (always auto / always ask / use threshold)
v1.1.0 introduces Cross-Project References — pull in context from any shared project in your organization directly inside a conversation or deep dive session. No more copy-pasting files. No more context switching. Bob sees the actual code from another project and uses it to inform his responses in real time.
Built for the moment when you're debugging a problem in one project and you know you solved something similar somewhere else — or when you want Bob to match a pattern, style, or architecture from a reference codebase.
📎 Referencing /payments-api...
Your organization's admin shares projects as named references in Bob's Workshop. Each shared project gets a custom alias and per-user access permissions. Once a project is shared, any permitted user can invoke it from the terminal by typing the alias as a slash command.
Bob receives the most semantically relevant files from the referenced project — summaries and full code content — and uses them to answer your question. He acknowledges the reference transparently so you always know where the context came from.
Type the alias directly in your message:
/payments-api How does the webhook retry logic work here?
Bob performs a vector search against the referenced project, retrieves the most relevant files, and responds with full cross-project awareness.
Type /ref inside any chat or deep dive session to see a numbered list of projects you have access to:
╔══════════════════════════════════════════════════════════╗
║ 📎 AVAILABLE REFERENCES ║
╠══════════════════════════════════════════════════════════╣
║ 1. /payments-api seedling/payments-service ║
║ 2. /mobile-app seedling/mobile ║
║ 3. /bob-workshop seedling/bobs-workshop ║
╠══════════════════════════════════════════════════════════╣
║ Enter number to select, 0 to cancel ║
╚══════════════════════════════════════════════════════════╝
Select a project by number. Choose whole project (broad vector search) or drill into a specific file.
When you need a specific file rather than a semantic search:
/payments-api — whole project (p) or specific file (f)? f
Filter files (or Enter for all): webhook
1. src/handlers/webhook_handler.ts
2. src/services/webhook_retry_service.ts
Select file: 1
Bob reads that exact file and responds with precise, file-specific knowledge.
Pin a reference so it applies to every message automatically — no need to type the alias each time:
/pin
The prompt updates to show the active sticky:
📌 /payments-api ›
Every message you send automatically includes context from the pinned project. Toggle it off at any time:
/pin # toggles off if already active
Deep dives have their own independent sticky reference. Each sandbox session can reference a completely different project — two deep dives in the same conversation can point to two different codebases simultaneously.
/ref Browse available reference projects
/pin Toggle sticky reference on or off
Cross-Project References require a Bob's Workshop organization account with shared projects configured by your admin. Platform tier required.
Conversation
bob chat "question" # AI coding partner
bob consult "question" # Strategic advice
bob conversations # List conversations
bob fork "topic" # Branch conversation
bob deepdive # Sandboxed exploration
Project Tools
bob index # Index codebase
bob analyse # Code review
bob analyse --auto # Auto-fix
bob autonomy # Full autonomous repair
bob push "message" # Git push
VaultBob — Backup & Restore
bob backup create # Back up project context
bob backup create --source # Back up source code
bob backup create --source --file # Back up one file
bob backup create --global # Full machine backup (Grid)
bob backup create --archive "name" # Named checkpoint
bob backup list # View revision history
bob backup restore # Interactive restore
The Crew — Multi-Agent Orchestration
bob agent spawn <name> "<task>" # Spawn a named agent
bob agent spawn <name> --persona # Spawn with a persona
bob agent list # List all agents
bob agent status # Detailed agent status
bob agent hub # Interactive command center
bob agent chat <name> # Chat with a specific agent
bob agent personas # List available personas
bob agent stop <name> # Stop an agent
bob agent reset <name> # Reset an agent
bob agent-run "mission" # Launch autonomous mission
bob agent-run --dry-run "mission" # Preview task map
bob agent-run --resume # Resume paused mission
Digital Twin
bob userbob "mission" # Launch digital twin simulation
bob command-center # Autonomous task board
bob cc --stream # Live decision stream
Profile & Identity
bob profile --cloud # Generate DNA profile
bob profile # View dashboard
bob byok set google <key> # Add BYOK key
Remote (SovereignLink)
bob serve # Start SovereignLink
bob remote chat "msg" # Remote execution
bob remote backup # Remote backup
bob remote restore # Remote restore
Configuration
bob login # Authenticate
bob whoami # Status
📖 Full command reference: https://seedling-io.gitbook.io/bob-cli/bobs-cli-product-wiki-and-user-guide/command-reference
Powered by the Frank Reasoning Engine. Bob learns how you work and adapts:
bob profile --cloud
bob chat --personalized "what should I focus on?"
Tier 1 — Local (Free) Tier 3 — Platform (Subscription)
───────────────────────── ─────────────────────────────────
▸ Your model (Ollama) ▸ Claude / Gemini
▸ Files on your machine ▸ Conversations sync to web
▸ Local profiling ▸ Cloud profiling + Frank Engine
▸ Local UserBob simulation ▸ UserBob + autonomous dispatch
▸ The Crew (local agents) ▸ Deep dives, forks, remote exec
▸ VaultBob backup & restore ▸ VaultBob + team license mgmt
▸ Zero cost ▸ Cross-project references
▸ Scales to enterprise
Same commands. Scale without changing tools.
conversationId is now stored in each project's ~/.bob/projects/{name}/project.json instead of the global config. When you cd between projects, bob chat automatically resumes the correct conversation for that project. No more cross-contamination between codebases.conversationId exists yet, Bob falls back to the global config value seamlessly. The new value is written to project.json on first interaction.project.json now tracks lastActive — Every time a conversation ID is written to a project, the lastActive timestamp is updated. Useful for future project-aware features.remote.ts conversation ID — bob remote now reads the active conversation ID from project scope, not the global config. Connecting to an Active Bob correctly reflects the current project's conversation.serve.ts conversation ID — bob serve now resolves the conversation ID from project scope before registering the daemon session, ensuring SovereignLink binds to the correct project conversation./alias, browse with /ref, or lock a project for the entire session with /pin. Each deep dive maintains its own independent sticky reference. Bob receives full file content — summaries and source code — and responds with real cross-project awareness. Organization account required./ref — Interactive reference browser. Browse all projects your org admin has shared with you, pick by number, choose whole-project vector search or drill into a specific file./pin — Sticky reference toggle. Lock a project for the session so every message automatically carries cross-project context without repeating the alias.bob agent-run — One command to launch a supervised autonomous mission against your real codebase.bob agent-run --dry-run — Preview DirectorBob's full task map before executing.bob agent hub — Interactive command center for your full agent team.bob backup create --source — Back up your actual code files with gitignore respect. Encrypted before leaving your machine. Independent of GitHub.bob backup create --source --file <path> — Surgical single-file backup and restore. Back up the exact file you're about to edit. Restore it instantly if something goes wrong.bob backup create --global — Full machine migration. One command captures your entire engineering history across all projects. Restore to any machine in minutes. Grid plan required.bob remote backup.bob remote index — SovereignLink now executes full project indexing remotely.bob remote analyse — SovereignLink now executes full QA analysis remotely.Bob's CLI exists because we believe the future of software development should be owned by the developer, not rented from a corporation.
The cloud is optional. The power is not.
This is AI for the people who build the future — not the companies who gatekeep it.
The AI coding tool that learns how you think. The only backup system that protects your engineering brain. The only terminal with an engineering department built in.
Sovereign. Free. Yours.
Bob's CLI · The Crew™ · VaultBob™ · SovereignLink™ · Bob's Workshop · Seedling
Written by Bob.
FAQs
Bob's CLI — AI coding assistant and Forge orchestrator
The npm package @bobsworkshop/cli receives a total of 215 weekly downloads. As such, @bobsworkshop/cli popularity was classified as not popular.
We found that @bobsworkshop/cli 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.
Security News
PolinRider expands across npm, Packagist, Go modules, and Chrome extensions, using hidden loaders to target developer environments.
Security News
Open source attacks are accelerating as AI coding agents pull in dependencies faster, with less human review.
Research
/Security News
Malicious Chrome and Firefox extensions posed as free VPNs while stealing clipboard data through later extension updates.