Use ModOnda in ChatGPT and Claude
ModOnda can run inside ChatGPT, Claude and Claude Code as a connector. You ask for a model in plain words; the assistant looks up library components, writes the Modelica, checks it, simulates it on ModOnda's servers and shows you an interactive plot in the chat.
This is separate from the workbench's own ✦ Assistant. The connector needs no OpenRouter key, and the models it saves are kept for one day, not in your browser.
Connect
The connector address is:
https://modonda.com/mcp- Claude (claude.ai, desktop, mobile): add a custom connector in Claude's connector settings and paste the address above. Claude opens ModOnda's sign-in page; choose Sign in with GitHub, then Allow.
- Claude Code:
claude mcp add --transport http modonda https://modonda.com/mcp, then run/mcpinside Claude Code and sign in. - ChatGPT: turn on developer mode in ChatGPT's settings, create a connector (app) with the address above and choose OAuth sign-in. Developer mode is needed until ModOnda is listed in the app directory.
Every connect asks you to sign in on modonda.com and shows which app is asking. The connection is yours alone; other people who add the connector sign in as themselves.
What to ask
- "What components does the Modelica Standard Library have for a DC motor with a gearbox?"
- "Build a mass-spring-damper with a 10 kg mass and simulate 5 seconds. Plot the position."
- "Simulate
Modelica.Blocks.Examples.PID_Controllerand tell me the overshoot." - "Is this model stable? Linearize it and show the eigenvalues." (paste a model or describe one)
- "My simulation is slow. Explain the model's structure and suggest a solver."
What the connector can do
| Tool | What it does |
|---|---|
list_classes, get_class |
Browse the ModOnda library and the Modelica Standard Library, and read any class. |
save_model, list_models |
Save a Modelica file in your temporary workspace (kept for a day) and list what is there. |
check_model |
Translate a model and report its balance, states, loops and warnings. |
simulate_model |
Simulate and return a summary plus an interactive plot: hover for values, click a variable to hide it, switch to separate axes. |
physics_check |
Catch missing or impossible parameters and blow-ups before you trust a result. |
linearize_model, model_structure |
Linear state-space model with eigenvalues, and why a model is stiff or slow. |
Models that read files or call external code cannot run on the server. Tools never change anything outside your temporary workspace.
Limits
Each account has a daily allowance for calls and simulation time, and one run at a time. When it is used up, the assistant tells you when it resets. Runs that take longer than the per-call limit are stopped; ask for a shorter stop time or a simpler model. The workbench in your browser has no such limits. Any library model the assistant names opens there at https://modonda.com/#<model name>, for example https://modonda.com/#Modelica.Blocks.Examples.PID_Controller; to take a model you built in the chat, ask the assistant for the complete file and paste it into the workbench.
Your data
The connector sends the text of the models you ask the assistant to save or simulate to ModOnda's server, where it is kept in memory for a day and then dropped. See the privacy notice. To disconnect, open your account page and remove the app under Connected apps; its access stops at once. Removing the connector in ChatGPT or Claude does not revoke it on our side, so do both.