The engine. On your hardware. On your terms.

$ git clone https://github.com/NovoMCP/novomcp \
    && cd novomcp/orchestrator \
    && pip install -r requirements.txt \
    && python main_https.py

$ git clone https://github.com/NovoMCP/novomcp \
    && cd novomcp \
    && docker compose up

Two install paths today (git clone + docker compose). The pip install novomcp and npx novo packages are name reservations for now; installable releases will land on docs.novomcp.com and in GitHub Releases. For the cheminformatics on its own, the lightweight version, novomcp-lite, is on PyPI today (pip install novomcp-lite): the same RDKit tools and public-API search as a two-dependency library or MCP server. The same code the engine runs on, packaged standalone.

$ curl -X POST http://localhost:8018/mcp/call \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer local' \
  -d '{
    "name": "get_molecule_profile",
    "arguments": { "smiles": "CC(=O)Oc1ccccc1C(=O)O" }
  }'

Local calls use the placeholder bearer token local. Wire NOVOMCP_API_KEY when you deploy behind an auth gate.

Eleven tools work the second the server boots.

Point env vars at your own services. Each capability speaks HTTP to any provider.

The engine speaks HTTP to each service. Ship the reference containers from the same repo, or swap in your own. The interfaces below stay open to any provider. The GPU services also ship as standalone open repos, deployable on their own. gromacs-md (MD), novomcp-nnp (neural-network potentials), and novomcp-qm (quantum chemistry). Compliance (check_compliance) is the exception. It calls the hosted FAVES API directly, which is a specific service with its own contract.

Observability built into the engine from the first stage.

The engine writes an immutable audit trail; compliance runs at the point of decision; memory persists across runs; agents acting on your pipeline are scoped and logged.