Interpret the results
It reads the drag and lift coefficients, the forces, the forces per part and the result images, and explains where the drag or downforce comes from.
AirShaper has an MCP server. MCP (Model Context Protocol) is the standard way AI assistants such as Claude and ChatGPT connect to other tools. Connect yours once and it can set up, run and read your simulations for you.
Have an AirShaper account? Connect your assistant

Upload wing_v2.stl, run it at 120 km/h and tell me the drag and lift.
Compare my last three spoiler variants. Which one has the least drag?
Which parts of the car produce the most drag in my latest simulation?
Lower the rear wing by 20 mm in my CAD model, run it again and compare it with the baseline.Needs a CAD connection as well
You ask, your assistant picks the steps. Setting up a simulation, running it, reading the results and comparing variants are all within its reach.
It reads the drag and lift coefficients, the forces, the forces per part and the result images, and explains where the drag or downforce comes from.
Add your CAD tool's MCP server to the same assistant, for example one for Onshape or Blender, and it can change the model itself and export the new geometry.
The CAD connection is a separate one that you add next to AirShaper; it is not part of our server.
It uploads each new version (STL, OBJ, STEP or IGES) and copies the settings from your baseline, so every variant runs the same way.
Diagnose, change, run, compare with the baseline and repeat, with several variants running at the same time.
You control the credits. Your assistant gets the cost first, and a run only starts once that exact amount is confirmed.
Let your AI explore and try, and let AirShaper do the aerodynamics: a wind tunnel validated solver that treats every variant the same way.
The case setup is consistent from run to run, so a difference between two variants comes from the design, not from the setup.
The mesh is built and refined automatically, with no mesh settings to manage.
Each simulation runs on its own machine, so a batch of variants does not wait in line.
These sessions were recorded with Claude. ChatGPT, Cursor and any other assistant that supports MCP connect to AirShaper the same way.
The assistant uses the Onshape and AirShaper connections together to reduce the drag of the Ahmed body, a standard test shape for car aerodynamics.
The assistant reads the results of a sports car, explains where the lift and drag come from, and prepares three variants to test.
The Design Advice map on the same Ahmed body shows where the surface should move in or out to reduce drag.
Today your assistant can order a Design Advice run and look at its images. Sending it the map itself, as data it can work with, is coming.
Add the server to your assistant and sign in. There is no token to copy.
https://app.airshaper.com/mcpCreating and running simulations needs an active subscription. On a free account, your assistant can only read the coefficients and preview images of your free runs, during each run’s first day. See the plans.
You can see and disconnect your assistants at any time in AirShaper under Settings, API & Integrations.
On the web or in the desktop app, go to Customize, Connectors, Add, then Add custom connector, and paste the URL. If asked for an OAuth client, choose “Use Claude’s published identity”. Once added, it is there in the mobile apps too. On Team and Enterprise plans, an owner adds it for the organization first.
Add a custom MCP server with the URL and choose OAuth as the authentication. ChatGPT calls this developer mode; which plans have it and where it sits in the settings is up to OpenAI.
Run the command below, then type /mcp in Claude Code and choose airshaper to sign in.
claude mcp add --transport http airshaper https://app.airshaper.com/mcpAdd the server to .cursor/mcp.json in your project, or to ~/.cursor/mcp.json for all projects. Cursor opens the sign-in the first time it connects.
{ "mcpServers": { "airshaper": { "url": "https://app.airshaper.com/mcp" } } }Add a remote (HTTP) MCP server with the URL, then sign in when it opens AirShaper. In VS Code: Command Palette, “MCP: Add Server”, HTTP.
Working from an assistant is made for subscribers. When something is not in your plan, your assistant is told which plan includes it.
On a free account? Your assistant can only read the coefficients and preview images of your free runs, during each run’s first day. Creating and running simulations needs a subscription.
See the plansThe server gives your assistant 17 tools. You never need to name them: your assistant picks the ones it needs.
Connecting a script, a CI job or your own agent? Create an MCP token in AirShaper under Settings, API & Integrations and send it as a bearer token.