Open-source workflow builder with cost controls and human oversight. Build AI automations with tools like Gmail, GitHub, and Supabase - with built-in spend caps and approval gates.

git clone https://github.com/seer-engg/seer
cd seer
cp .env.example .env
# Add your OPENAI_API_KEY to .env
docker compose up

That's it. Open http://localhost:5173 and start building workflows.

No accounts, no sign-ups — the app runs in local mode by default with no login required.

docker compose up starts everything:

API docs are at http://localhost:8000/docs.

Create .env from the example:

cp .env.example .env

Required:

OPENAI_API_KEY=sk-...       # Or any LLM provider key

Optional integrations (add as needed):

ANTHROPIC_API_KEY=sk-ant-... # Claude models
OPENROUTER_API_KEY=sk-or-... # Kimi, other models
GOOGLE_CLIENT_ID=...         # Google integrations (Gmail, Sheets)
GOOGLE_CLIENT_SECRET=...
BRAVE_SEARCH_API_KEY=...     # Web search tool

Docker automatically configures DATABASE_URL and REDIS_URL.

Auth mode is controlled by a single env var:

For clerk mode, also set CLERK_JWKS_URL, CLERK_ISSUER, CLERK_SECRET_KEY, and VITE_CLERK_PUBLISHABLE_KEY. See .env.example for details.

If you prefer running services directly:

# Start only infra
docker compose up postgres valkey browserless -d

# Backend (requires Python 3.12+, uv)
uv sync
uv run aerich upgrade
uv run uvicorn seer.api.main:app --reload --port 8000

# Worker
uv run taskiq worker seer.worker.broker:broker

# Frontend (requires Node 22+)
cd seer-frontend
npm install
npm run dev

seer/
├── src/seer/              # Python backend
│   ├── api/               # FastAPI routers, middleware
│   ├── services/          # Business logic (workflows, triggers, integrations)
│   ├── core/              # Workflow compiler/runtime, schema models
│   ├── tools/             # Tool registry, executor, provider implementations
│   ├── agents/            # LangGraph-based agent orchestration
│   ├── worker/            # Taskiq background tasks, trigger polling
│   ├── database/          # Tortoise ORM models/config
│   └── auth/              # Auth provider abstraction (Clerk / local)
├── seer-frontend/         # React + TypeScript + Vite frontend
│   └── src/
├── migrations/            # Database migrations
├── tests/                 # Backend tests
├── scripts/               # Maintenance scripts
└── docker-compose.yml     # Full stack setup

Run migrations after pulling updates:

# Docker
docker compose exec api uv run aerich upgrade

# Local
uv run aerich upgrade

docker compose up              # Start all services
docker compose up -d           # Start in background
docker compose logs -f api     # Follow API logs
docker compose down            # Stop all services
docker compose down -v         # Stop and remove data
uv run pytest                  # Run backend tests

• Cost Controls - Per-workflow spend caps and token limits prevent runaway AI expenses
• Human Oversight - Approval gates for critical operations; complete audit trails
• Powerful Integrations - Gmail, GitHub, Supabase, PostgreSQL with minimal permissions
• AI-Native - Chat interface for workflow design; intelligent error handling
• Self-Hostable - git clone + docker compose up. No accounts needed

Complete Documentation - Comprehensive guides, architecture, and examples on DeepWiki

AGPL-3.0 License - See LICENSE.