GROMACS MCP Server

Drive GROMACS molecular-dynamics simulations and analysis from Claude through a self-contained Docker container that ships GROMACS 2025.4, with async job tracking for long-running runs.

   
Type MCP server
Supplier MacromNex
Availability Alpha (repo created 2025-12-24, last updated 2026-04-17)
Pricing Free / OSS (Unverified — upstream declares no repository LICENSE file; the README lists only “LGPL (GROMACS)”, which is GROMACS’s own license, not the wrapper’s. Do not assume redistribution terms until a license is published upstream)
Capabilities Read/Write — runs GROMACS commands and simulations against files in the mounted working directory
Verified works · 2026-07-20
Security caution · 2026-07-20 — provenance matches MacromNex, no repo LICENSE, alpha, no OSV advisories

How to install

Both paths register the server with Claude Code over stdio; Claude launches the docker run process itself on demand (no separate long-running terminal). The container mounts your current directory so GROMACS can read/write your input and output files.

  • PrerequisiteDocker installed and running, plus Claude Code.

  • Claude Code — pull the prebuilt image (recommended):
    docker pull ghcr.io/macromnex/gromacs_mcp:latest
    claude mcp add gromacs -- docker run -i --rm --user `id -u`:`id -g` -v `pwd`:`pwd` ghcr.io/macromnex/gromacs_mcp:latest
    

    (Run the claude mcp add command from the directory that holds your simulation files — the -v `pwd`:`pwd` mount and `id -u`:`id -g` are evaluated by your shell at registration time, so the container sees that directory at the same absolute path.)

  • Claude Code — build locally:
    git clone https://github.com/MacromNex/gromacs_mcp.git
    cd gromacs_mcp
    docker build -t gromacs_mcp:latest .
    claude mcp add gromacs -- docker run -i --rm --user `id -u`:`id -g` -v `pwd`:`pwd` gromacs_mcp:latest
    
  • Claude Desktop — Claude Desktop cannot expand shell backticks in a JSON config, so it cannot reproduce the `id -u`/`pwd` substitutions the way Claude Code does. Registration not documented upstream for Claude Desktop — the Desktop claude_desktop_config.json entry would need the docker run args pre-expanded to literal values, e.g.:
    {
      "mcpServers": {
        "gromacs": {
          "command": "docker",
          "args": ["run", "-i", "--rm", "-v", "/absolute/path/to/your/workdir:/absolute/path/to/your/workdir", "ghcr.io/macromnex/gromacs_mcp:latest"]
        }
      }
    }
    

    (replace /absolute/path/to/your/workdir with the absolute path of the directory holding your GROMACS files — e.g. /Users/you/md-run; adapt/verify against the upstream README before relying on it.)

What it does

Exposes six tools that front the GROMACS command-line toolkit inside the container:

  • run_gromacs_command — execute a single GROMACS command with input/output files and parameters; also analyzes TPR files.
  • run_gromacs_workflow — orchestrate a predefined multi-step workflow (a demo workflow is documented; other named workflows are not enumerated upstream).
  • submit_md_simulation — submit a long-running MD production run (input TPR, step count, output dir, job name) for asynchronous background execution.
  • submit_batch_analysis — batch-analyze multiple molecular systems / TPR files in parallel.
  • get_job_status — poll a background job’s progress without pulling full results.
  • get_job_result — retrieve a completed job’s outputs.

Operations returning in under ~10 minutes run synchronously; longer tasks use the async job-submission tools (submit_*get_job_statusget_job_result).

Primary use cases: run and monitor GROMACS MD simulations from Claude, batch trajectory/energetics analysis across multiple systems, TPR inspection.

Notes

Alpha-stage community project, not vendor-affiliated. The container bundles GROMACS 2025.4, so no local GROMACS install is needed — everything runs inside Docker. GPU-acceleration support, memory requirements, and the full set of named workflows / supported force fields are not documented upstream (detail.md covers only tool signatures); verify capability against your own systems. Force-field files and topology preparation are your responsibility — the tools operate on files you provide in the mounted directory. Because the server writes into the mounted working directory, treat it as Read/Write and mount only the directory you intend GROMACS to touch. For MD runners without Docker, see the catalogued OpenMM MCP; for GROMACS+VMD visualization workflows, egtai/gmx-vmd-mcp remains deferred pending a followable install path.

Sources


Installed this tool?

Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.