CrewDock

Open-source control plane for operating focused AI agents. CrewDock gives teams one self-hosted place to manage agent roles, task state, approvals, activity streams, knowledge context, and cost visibility.

Built by Felhen, an applied AI company building governed AI operating systems for real businesses.

Why CrewDock?

Most teams start with a powerful agent in a terminal. That works until the work needs coordination, memory, approvals, observability, and a clear operating model. CrewDock is the open-source base layer for that next step.

• Focused agents — create agents with explicit roles, models, prompts, and responsibilities
• Operational surface — chat, tasks, approvals, schedules, activity, and costs in one place
• Knowledge-aware execution — agents search your document base before responding
• Governed autonomy — human-in-the-loop workflows and audit trails for agent actions
• Self-hosted by default — Docker Compose, portable stack, no vendor lock-in

Features

Core

• Multi-agent management — create agents with custom system prompts, models, and roles
• Agent chat with streaming — real-time token-by-token responses via SSE
• Knowledge base integration — agents search your documents (QMD) before responding
• Task orchestration — Kanban board with drag-and-drop, cron scheduling, state machine
• Real cost tracking — actual token usage recorded from API calls
• Agent templates — 10 pre-configured agents across 6 categories, one-click install

Platform

• JWT authentication — email/password login with setup flow for first user
• Task scheduler — APScheduler executes recurring tasks via LLM
• Persistent chat history — Redis-backed with in-memory fallback
• Activity feed — real-time event stream via Server-Sent Events
• Approval workflows — human-in-the-loop for agent actions
• Webhook notifications — HMAC-signed event delivery
• Plugin system — extensible architecture with lifecycle management
• Gateway-agnostic — abstract adapter supports any agent runtime

Design

• Branded theme — cyan/maritime palette, not generic gray
• Dark mode — navy-tinted premium dark theme
• Mobile responsive — hamburger menu, responsive grids
• Onboarding — guided setup for new users

Built by Felhen

CrewDock is public because the basic shape of agent operations should be inspectable: roles, context, task state, approvals, observability, and cost control. Felhen uses this foundation to build private AI operating systems, company knowledge workflows, and governed agent operations for companies that need AI to move real work, not just answer prompts.

• Use CrewDock if you want a practical self-hosted agent control plane.
• Contribute if you are building agent runtime, knowledge, approval, or observability primitives.
• Talk to Felhen if your company wants help designing or deploying governed AI operations.

Stack

Layer	Technology
Backend	Python 3.12, FastAPI, SQLAlchemy 2.0 (async), Alembic, Pydantic v2
Frontend	Next.js, React 19, TypeScript, Tailwind CSS, shadcn/ui
Database	PostgreSQL 16
Cache	Redis 7
LLM	Anthropic API (Claude Opus / Sonnet)
Auth	JWT + static bearer token (backwards compatible)
Deploy	Docker Compose
CI	GitHub Actions (ruff, mypy, pytest, ESLint, Next.js build)

Quick Start

Docker Compose (recommended)

# Clone
git clone https://github.com/antonio-mello-ai/crewdock.git
cd crewdock

# Configure (generate secrets automatically)
cp .env.example .env
sed -i "s/changeme_password/$(openssl rand -hex 24)/" .env
sed -i "s/changeme_token/$(openssl rand -hex 32)/" .env

# Add your Anthropic API key for agent chat
# Get one at https://console.anthropic.com
echo "ANTHROPIC_API_KEY=sk-ant-..." >> .env

# Start
docker compose up -d

# Create database tables
docker compose exec backend alembic upgrade head

# Open the dashboard
open http://localhost:3001

Development

# Start infrastructure only
docker compose -f compose.dev.yml up -d

# Backend
cd backend
python3.12 -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
uvicorn app.main:app --reload --port 8001

# Frontend (in another terminal)
cd frontend
npm install
npm run dev

API

The backend exposes a RESTful API at /api/v1/:

Endpoint	Description
GET /health	Health check
POST /auth/login	JWT authentication
POST /auth/setup	Create first admin user
GET/POST /agents	Agent management
POST /chat/{id}	Chat with agent (non-streaming)
POST /chat/{id}/stream	Chat with agent (SSE streaming)
GET/POST /tasks	Task orchestration
GET /activity	Activity feed
GET /costs	Cost tracking
POST /knowledge/search	Knowledge base search
GET /events	Server-Sent Events stream
GET /scheduler/jobs	List scheduled tasks
GET /gateway/status	Gateway connection status
GET/POST /skills	Skill registry
GET/POST /approvals	Approval workflows
GET/POST /webhooks	Webhook management
GET/POST /plugins	Plugin management

Full OpenAPI spec available at /openapi.json (29 endpoints).

Architecture

┌─────────────────────────────────────────────┐
│              Caddy (reverse proxy)          │
├──────────────────┬──────────────────────────┤
│  Frontend        │   Backend (FastAPI)      │
│  Next.js :3001   │   :8001                  │
├──────────────────┴──────────────────────────┤
│         Anthropic API (Claude LLM)          │
├─────────────────────────────────────────────┤
│   PostgreSQL + Redis + QMD Knowledge Base   │
├─────────────────────────────────────────────┤
│      Gateway Adapter + Plugin System        │
└─────────────────────────────────────────────┘

Project Structure

crewdock/
├── backend/           # FastAPI application
│   ├── app/
│   │   ├── api/       # 15 route handlers
│   │   ├── core/      # Config, auth, database, security
│   │   ├── models/    # 10 SQLAlchemy models
│   │   ├── schemas/   # Pydantic request/response schemas
│   │   ├── services/  # LLM, scheduler, knowledge, cost tracker
│   │   └── plugins/   # Plugin system
│   ├── alembic/       # Database migrations
│   └── tests/         # 14 pytest tests
├── frontend/          # Next.js application
│   └── src/
│       ├── app/       # 11 pages + chat + login
│       ├── components/# Atoms, forms, organisms
│       ├── hooks/     # React Query hooks
│       └── lib/       # API client, auth, templates
├── website/           # Landing page (crewdock.ai)
├── docs/              # Roadmap, changelog, benchmarks
├── compose.yml        # Production Docker Compose
└── .github/workflows/ # CI pipeline

Links

• Website: crewdock.ai
• Felhen: felhen.ai
• Questions and ideas: GitHub Discussions
• Issues: GitHub Issues
• Roadmap: docs/roadmap.md
• Changelog: CHANGELOG.md
• Known Issues: docs/known-issues.md
• Document Metadata: docs/document-metadata.md
• Support: SUPPORT.md
• Contributing: CONTRIBUTING.md

License

MIT