comsol-mcp

comsol-mcp

MCP server for visible COMSOL automation that attaches to a running COMSOL Multiphysics Server, enabling shared model state with the Desktop GUI for collaborative modeling.

Category
Visit Server

README

COMSOL MCP

AI-driven MCP server for COMSOL with a visible Desktop client workflow.

comsol-mcp is built for the workflow where:

  • you manually start COMSOL Multiphysics Server
  • COMSOL Desktop connects to that server as a visual client
  • an MCP client attaches to the same server-side model
  • modeling changes remain visible instead of running as a black-box batch job

This repository is the new mainline for visible COMSOL automation in this codebase. The old batch-oriented route is now a legacy reference only.

Why This Project Exists

Many COMSOL automation flows are powerful but opaque. They can compile Java, run batch jobs, or drive generic server-side simulations, but they do not always let you watch the model evolve in the Desktop GUI.

comsol-mcp focuses on a different goal:

  • visible GUI changes
  • shared server-side model state
  • no Desktop-side polling bridge
  • no GUI automation layer
  • no black-box batch-only workflow

Core Workflow

This project uses an attach-first workflow:

  1. Start COMSOL Multiphysics Server manually.
  2. Note the real listening port from the server console.
  3. Connect COMSOL Desktop to that same port.
  4. Import the server-side model into Desktop if needed.
  5. Attach MCP to the same server with server_connect(host, port).
  6. Create or load the working model.
  7. Use MCP tools to modify the shared server-side model.
  8. Watch the same model update in COMSOL Desktop.

See the workflow guide for the full version.

Quick Start

Prerequisites

  • Windows with COMSOL installed locally
  • A valid COMSOL license
  • Python 3.10 or newer
  • A manually started COMSOL Multiphysics Server

Install

From source:

git clone <your-repo-url> comsol-mcp
cd comsol-mcp
python -m pip install -e .

Set environment variables:

$env:COMSOL_ROOT = "C:\Program Files\COMSOL\COMSOL63\Multiphysics"
$env:COMSOL_SERVER_MCP_HOME = "$PWD\comsol-server-home"

Start the MCP server:

python -m comsol_mcp.mcp_server

Or use the helper script:

.\scripts\start_comsol_mcp.ps1 -Python python -ComsolRoot "C:\Program Files\COMSOL\COMSOL63\Multiphysics" -McpHome "$PWD\comsol-server-home"

MCP Config Example

See:

Minimal Visible Demo

server_connect("localhost", <actual_port>)
model_create("VisibleServerModel")
ensure_component("comp1", 2)
ensure_geometry("comp1", "geom1", 2)
ensure_mesh("comp1", "mesh1")
create_feature("comp1", "geom1", "r1", "Rectangle", "[{\"name\":\"size\",\"values\":[\"60[mm]\",\"30[mm]\"]},{\"name\":\"pos\",\"values\":[\"-30[mm]\",\"-15[mm]\"]}]", true)
run_feature("mesh", "mesh1", "comp1")

Desktop should show the same geometry because Desktop and MCP share the same server-side model.

See examples/attach_first_demo.md for the same flow in document form.

Tool Surface

Current tools:

  • server_info()
  • server_start(...)
  • server_connect(host, port, model_name="")
  • server_disconnect(shutdown_server=false)
  • model_create(name="Server Model")
  • model_load(path)
  • model_tree()
  • get_parameters()
  • set_parameters(parameters_json)
  • ensure_component(component="comp1", dimension=2)
  • ensure_geometry(component="comp1", geometry="geom1", dimension=2)
  • ensure_mesh(component="comp1", mesh="mesh1")
  • create_feature(component, geometry, tag, feature_type, properties_json="[]", run_geometry=false)
  • update_feature(component, geometry, tag, properties_json, run_geometry=false)
  • delete_feature(component, geometry, tag, run_geometry=false)
  • run_feature(collection, tag, component="comp1")
  • run_study(study_tag="")
  • save_model(path="")

The modeling tool surface is intentionally stable. This repository cleanup does not rename or remove these tools.

Recommended Entry Point

The recommended entrypoint is:

server_connect("localhost", <actual_port>)

server_start() is still available, but it is now an advanced fallback. Use it only when you want MCP to own the COMSOL server lifecycle and accept that COMSOL may choose a different listening port automatically.

Known Limitations

  • COMSOL Desktop may not automatically display the server-side model after a successful connection; you may need to import or switch to the existing server model.
  • server_connect() can succeed while no current working model is selected; in that case use model_create() or model_load().
  • The server console output and the actual listening port should always be verified against the real active listener.
  • Desktop graphics may require a light refresh after server-side model changes.

See the troubleshooting guide.

How This Differs From Other MCPs

Versus abaqus-mcp-server

abaqus-mcp-server is a GUI scripting / GUI automation style project. It works with an already running GUI and uses GUI automation techniques to trigger actions.

comsol-mcp does not use pywinauto-style GUI automation for COMSOL. Instead, it attaches directly to COMSOL Multiphysics Server and drives the same server-side model that Desktop visualizes.

Versus generic COMSOL automation MCPs

Public COMSOL MCP listings often emphasize solver breadth, mesh generation, physics setup, and parametric sweep coverage.

This project emphasizes a different value proposition:

  • attach-first
  • visible Desktop workflow
  • non-black-box modeling
  • shared model state between Desktop and MCP

It is not just "another COMSOL automation MCP"; it is specifically about making the workflow visible and collaborative with the Desktop client.

Versus the old batch route

The old batch route used:

  • comsolcompile
  • comsolbatch

That route is still useful for offline jobs, but it is no longer the public identity of this project. In this repository, it is treated as a legacy reference rather than the mainline.

See docs/differences.md for a fuller comparison.

Legacy / Previous Batch Route

An earlier internal comsol-mcp route in this codebase was batch-oriented and wrapped comsolcompile plus comsolbatch. That legacy route is intentionally not the public face of this repository.

This repository focuses on:

  • COMSOL Multiphysics Server
  • Desktop as a visual client
  • MCP attach to the same model

Attribution

  • Built on top of MPh
  • Inspired by COMSOL client-server workflows and related MCP explorations
  • Not affiliated with COMSOL
  • Does not include COMSOL binaries or proprietary assets

License

MIT. See LICENSE.

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured