openterms-mcp

Openterms Open source cryptographic consent receipts + programmable guardrails + provider verification for AI agents. Your agent proves what it agreed to. Your policy controls what it's allowed to do. The API provider can verify both. License Tests Live Demo · Quickstart · Self-Host · MCP Server · Contributing · Open Receipt Spec What It Does Openterms sits between your AI agent and the actions it takes. Three layers: 1. Receipts — Before your agent calls an API, it gets an Ed25519-signed receipt. Canonical JSON (RFC 8785), SHA-256 hash, real cryptography. Anyone can verify it using public keys — no API key needed, no trust in the server required. 2. Policy Engine — Daily spending caps, action type whitelists, escalation thresholds. The policy engine evaluates before the receipt is signed. Denied actions never get a receipt. 3. Provider Verification — API providers register their terms URL and verify agent consent before serving requests. One public GET call. Both sides of the transaction trust the proof. Quickstart 60 seconds to your first receipt: git clone https://github.com/jstibal/openterms.git cd openterms pip install flask pyjwt cryptography pyyaml bash quickstart.sh Or with Docker: git clone https://github.com/jstibal/openterms.git cd openterms docker compose up --build # In another terminal: bash quickstart.sh Receipt issuance is free — no wallet, no deposit, no payment required. Self-Host Run your own Openterms instance: # Option 1: Direct pip install flask pyjwt cryptography pyyaml python run.py # Server at http://localhost:5000 # Option 2: Docker docker compose up --build # Option 3: Point MCP server at your instance export OPENTERMS_API_URL=http://localhost:5000 python openterms_mcp_server.py Everything runs locally. SQLite database, no external dependencies. Hosted Service Don't want to self-host? Use the hosted instance at openterms.com — same open source code, managed for you. MCP Server 10 tools for AI agents: Tool What it does is…