Open-Source Multi-Agent Workflow Orchestrator
Design agent workflows visually. Run them reliably. Monitor everything.
AgentForge is a self-hosted platform that enables developers and teams to visually design, deploy, execute, and monitor multi-agent AI workflows through a drag-and-drop DAG editor, real-time execution tracing, cost tracking, and a marketplace for reusable agent templates.
- Visual DAG Editor — Drag-and-drop canvas for composing multi-agent workflows with 7 node types (Agent, Tool, Router, Evaluator, Human-in-the-Loop, Input, Output)
- LangGraph-Powered Execution — Workflows compile to LangGraph state graphs for reliable, stateful execution with checkpointing
- Real-Time Monitoring — Watch execution animate on the DAG canvas with live node status, data flow, and timing
- Cost Tracking — Per-node token counting, per-workflow cost aggregation, and budget alerts
- Agent Template Registry — Browse, install, and publish reusable agent templates
- MCP Server Integration — Connect any MCP server as a tool for your agents
- API & Webhook Triggers — Expose workflows as REST endpoints or trigger via webhooks
- RBAC — Team workspaces with role-based access control (Owner, Admin, Editor, Viewer)
| Layer | Technology |
|---|---|
| Frontend | Next.js 14, React 18, TypeScript, Tailwind CSS, React Flow |
| Backend | FastAPI, Python 3.12, SQLAlchemy 2.0, Alembic |
| Agent Framework | LangGraph |
| Protocols | MCP (Model Context Protocol) |
| Database | PostgreSQL 16 |
| Cache/Queue | Redis 7 |
| Storage | MinIO (S3-compatible) |
| Observability | OpenTelemetry, Langfuse |
| Deployment | Docker Compose |
- Docker Engine 24+ and Docker Compose v2
- 8 GB RAM minimum
- (Optional) LLM API keys for OpenAI, Anthropic, or Google
# 1. Clone the repository
git clone https://github.com/DeryFerd/agentforge.git
cd agentforge
# 2. Copy environment file and add your API keys
cp .env.example .env
# Edit .env to add OPENAI_API_KEY or other provider keys
# 3. Start all services
docker compose up -d
# 4. Run database migrations
docker compose exec api alembic upgrade head
# 5. Open your browser
# Frontend: http://localhost:3000
# API Docs: http://localhost:8000/docs
# Langfuse: http://localhost:3001# Backend
cd backend
pip install -e ".[dev]"
uvicorn app.main:app --reload
# Frontend
cd frontend
npm install
npm run dev┌─────────────┐ ┌─────────────┐ ┌──────────────┐
│ Next.js │───▶│ FastAPI │───▶│ LangGraph │
│ Frontend │◀───│ API Layer │◀───│ Engine │
└─────────────┘ └──────┬──────┘ └──────┬───────┘
│ │
┌─────┴─────┐ ┌─────┴─────┐
│ PostgreSQL│ │ Redis │
│ + MinIO │ │ (Queue) │
└───────────┘ └───────────┘
See STRUCTURE.md for the complete file-by-file architecture map.
| Document | Description |
|---|---|
| PRD.md | Product Requirements Document |
| system-design.md | Technical Architecture |
| STRUCTURE.md | Complete file structure + connection map |
| agents.md | Agent Behavior & Execution Model |
| plan.md | Implementation Plan |
We welcome contributions! See CONTRIBUTING.md for guidelines.
See plan.md for the current implementation plan and post-MVP roadmap.
Built for the AI orchestration era. AgentForge demonstrates mastery of multi-agent patterns, production engineering, and developer experience — all in one open-source platform.