Self-hosted, AI-native contract lifecycle management — your contracts never leave your servers.
🚀 Try the live app · ▶ Watch the 60-second demo · Quickstart · Contributing · Legal contributors
Enterprise contract AI costs six figures a year — and locks your data in someone else's cloud.
draftLegal does it all, open source, on your own servers: draft, review, redline, negotiate, approve, e-sign, and track obligations — with AI agents at every step of the lifecycle.
The market's two leaders prove the thesis: Harvey has the intelligence (analysis, drafting) but no execution layer; Ironclad has the execution (approvals, signatures, tracking) but weak intelligence — so they partnered. draftLegal builds both, natively, in one platform you control.
What makes it different: most CLM tools store your contracts. draftLegal reasons over them. Ask your whole portfolio a question — "A vendor quoted us $200k for an 8-week SOW. Is that fair?" — and it answers from your actual contract history, grounded and cited.
| 🤖 Agent-first | Seven specialist agents (review, draft, redline, approve, portfolio, ask, assist) on a LangGraph orchestrator — grounded in your contracts and cited so you can verify every answer. |
| 📑 Extract + cite | Drop in any contract; every clause, date, and dollar is extracted, risk-scored, and cited to the source page. |
| 📊 Portfolio intelligence | Hybrid retrieval (pgvector + Elasticsearch + RRF fusion) lets the agent reason across your entire contract portfolio — pricing benchmarks, exposure, what auto-renews. |
| ✍️ Redline & negotiate | Ask for counter-proposals; get conservative / moderate / aggressive options with exact replacement language, apply in a click. |
| ✅ Full lifecycle | Approval workflows, in-platform e-signature, and post-signature obligation tracking — "on rails." |
| 🔑 Bring your own model | Anthropic, OpenAI, or Google Gemini — configure your provider and key. No vendor lock-in. |
| 🔒 Self-hosted & private | docker compose up on your own infra. Your contracts never touch a third-party SaaS. |
Just want to look around? A hosted instance runs at app.draft-legal.com — no setup needed. Self-hosting (below) is the recommended way to run it on your own contracts.
Prerequisites: Docker (running), Node 22+ with pnpm 9+, and Python 3.11+.
git clone https://github.com/AniketTati/draft-legal.git
cd draft-legal
pnpm dev:setup # infra + deps + Python venv + DB migrate + seed (one command)
pnpm dev # starts web + api + agentsThen open http://localhost:5173 and log in with admin@demo.com / password123.
The setup script copies .env.example → .env for you. The app boots with no API key (browse, upload, manage contracts) — to enable AI features add at least one key to .env:
GOOGLE_API_KEY=... # generous free tier — easiest to start
# or ANTHROPIC_API_KEY / OPENAI_API_KEY…then restart pnpm dev. That's it — you have a private, AI-native CLM running locally.
Manual setup (if you'd rather run the steps yourself)
cp .env.example .env # ports already match docker-compose
docker compose up -d # Postgres, Redis, Elasticsearch, MinIO, Gotenberg
pnpm install
cd apps/agents && python3 -m venv .venv && ./.venv/bin/pip install -r requirements.txt && cd ../..
pnpm --filter api db:generate
pnpm --filter api db:migrate
pnpm --filter api db:seed # admin@demo.com / password123
pnpm --filter api exec tsx scripts/seed-ai-demo.ts seed # demo contracts
pnpm dev┌─ React + Vite (web :5173) ─────────────────────────────────┐
│ dashboard · contract editor · agent chat · lifecycle │
└───────────────┬────────────────────────────────────────────┘
│ REST + SSE
┌───────────────┴─ Fastify + Prisma (api :3001) ─────────────┐
│ auth · RBAC · contracts · approvals · signatures · obligations
│ hybrid retrieval (pgvector + Elasticsearch + RRF) │
└───────────────┬────────────────────────────────────────────┘
│ internal API
┌───────────────┴─ FastAPI + LangGraph (agents :8002) ───────┐
│ 7 agents · tool-calling · grounded RAG · BYO model │
└─────────────────────────────────────────────────────────────┘
Postgres(+pgvector) · Redis · Elasticsearch · MinIO · Gotenberg
| Layer | Tech |
|---|---|
| Frontend | React 18, Vite, Tailwind, shadcn/ui |
| API | Node 22, Fastify, Prisma, PostgreSQL + pgvector |
| Agents | Python 3.11, FastAPI, LangGraph |
| Retrieval | pgvector (dense) + Elasticsearch (BM25) + RRF fusion |
| AI providers | Anthropic · OpenAI · Google Gemini (switchable, BYOK) |
| Infra | Redis (BullMQ), MinIO (S3), Gotenberg (PDF) |
draftLegal is built in the open and we'd love your help. There are two tracks:
- 🧑💻 Developers → CONTRIBUTING.md. Start with a
good first issue. - ⚖️ Legal & contract experts → CONTRIBUTING-LEGAL.md. No code required — contribute clause libraries, playbook rules, and templates that make the AI smarter. This is the part no closed-source CLM can do.
New here? The best first step is to pnpm dev:setup, try the demo, and tell us what felt rough in a Discussion.
| Path | When |
|---|---|
| Docker Compose (above) | Dev machines, internal demos, evaluation |
| Cloud Run | First 1–10 users, public demo — see docs/operations/20-CLOUD-RUN-LAUNCH.md |
| Kubernetes / EKS | Production with paying customers — see docs/operations/19-DEPLOYMENT-STRATEGY.md |
The Dockerfiles are shared — the same images run on Compose, Cloud Run, EKS, Render, Fly, or any plain Docker host.
AGPL-3.0 — free to use, self-host, fork, and build on. If you run a modified version as a network service, the AGPL asks that you share your changes back, which keeps draftLegal open for everyone. Want to build a commercial product without the network-copyleft terms? A separate commercial license is available — get in touch.
Questions, commercial licensing, partnerships, or press — reach out: aniket.tatipamula@gmail.com
⭐ If this is useful, star the repo — it's how others find it.