Skip to content
← All news
Announcements

NovoMCP v1.0.0 — the engine is now open source

The computational chemistry engine that has been powering our autonomous drug-discovery and materials-science demos is now open source. Apache-2.0, MCP + REST, 67 tools, self-hostable. One git clone away.

July 22, 2026

Today NovoMCP goes open source. v1.0.0 is live on GitHub under Apache-2.0.

What it is

The computational chemistry engine that has been powering our autonomous drug-discovery and materials-science demos. It speaks both MCP (Model Context Protocol — reachable from Claude Desktop, Cursor, Codex, Zed, Cline, and any MCP-compatible AI assistant) and REST (curated OpenAPI 3.1 spec at api.novomcp.com/openapi.json, callable from any pipeline that posts JSON).

67 tools in the catalog. ~14 work with a fresh clone and no wiring. The rest unlock as you wire the optional compute services they need — docking, MD, QM, structure prediction, ADMET. tools/list hides everything that can't actually run; no "unavailable" errors on tools you haven't set up.

The one-line install

Python 3.10+:

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

Or with Docker Compose:

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

Engine on http://localhost:8018. Point your AI assistant at it and start calling tools.

Two domains, one engine

Drug discovery is the design center. An 11-stage autonomous funnel (agm <disease>), ADMET prediction, molecular docking (AutoDock-GPU), molecular dynamics (GROMACS + HMR), protein structure prediction (OpenFold3), clinical outcomes prediction (NovoExpert-3).

Materials science is covered too. First-class MCP prompts for OLED emitter library screening (HOMO/LUMO, S1/T1, TADF ranking) and battery electrolyte stability screening (voltage windows for Li-ion / Na-ion / aqueous). Direct Materials Project integration. Backed by xTB, CREST, MACE, ANI-2x, AIMNet2.

Two ways to run it

  1. Self-hosted — free forever. The complete engine (MCP + REST + 67 tools + all ADMET / property prediction models + all compute-service reference containers), Apache-2.0. You run it on your infrastructure.
  2. Cloud deployment — contact for pricing. NovoMCP operates a per-org deployment on your behalf. Dedicated subdomain, SAML SSO, GPU pool for docking + MD, backups + DR, named on-call. Includes the certified FAVES compliance API for regulated pipelines (SLA, immutable audit-log retention, IQ/OQ/PQ documentation).

Weekly release cadence starts now

  • v1.0.1 (next week) — pip install novomcp + prebuilt Docker on GHCR
  • v1.0.5 — Chrome extension + Word add-in polish
  • v1.1.x — Omics data pack (108K target-disease associations, SQLite)
  • v1.1.5 — self-hosted similarity index server
  • v1.2.7 — Materials autonomous mode (mgm: OLED / electrolyte / catalyst)
  • v1.5.x — NovoWorkbench desktop (macOS + Windows) as OSS

Full roadmap: docs.novomcp.com/product-roadmap.

Get involved

If you work in computational chemistry, ML for drug discovery, materials for batteries / OLEDs / catalysts, or you're teaching a class where students need a real engine — clone it and try it. agm aspirin on a fresh install shows you what's possible.

— Ari