Skip to content
Tresca

OpenMDAO MCP servers and AI agent skills

There is no general OpenMDAO MCP server. The one listing, pycycle-mcp from Carnegie Mellon's Design Research Collective, covers pyCycle engine-cycle models built on OpenMDAO: it manages cycles, updates variables, runs sweeps and evaluates total derivatives, and feeds thrust and fuel burn into a CPACS aircraft pipeline. The OpenMDAO team publishes no AI integration. For other OpenMDAO problems, an agent has to write and run the Python itself.

  • No official NASA server yet.
  • 1 MCP server listed.
NASA1 serverChecked 4 Nov 2026

Integrations for OpenMDAO

Community and commercial servers

MCP serverCommunity

pycycle-mcpby Design Research Collective, Carnegie Mellon University (cmudrc)

An MCP server for pyCycle engine-cycle models built on OpenMDAO: run cycles, sweep inputs and evaluate total derivatives.

LocalCan modify models2· updated 3 weeks ago

Listings are sorted by origin, then by recent activity. GitHub data is refreshed nightly. Nothing here is ranked.

How AI connects to OpenMDAO

An MCP server or skill can only do what OpenMDAO lets programs do. OpenMDAO exposes 2 automation interfaces:

  • OpenMDAO Python framework (Python): OpenMDAO is itself a Python library: you build models from explicit and implicit components grouped together, attach solvers and a driver, and let the framework compute total derivatives for gradient-based optimisation, which the project says scales to hundreds or thousands of design variables. Because models are ordinary Python, any agent that can run Python can build one, and a server like pycycle-mcp adds structure and validation on top. Documentation
  • pyCycle (Python): pyCycle is a thermodynamic cycle-modelling library, mainly for jet engine performance, maintained in the OpenMDAO GitHub organisation; release 4.2.0 needs OpenMDAO 3.10.0 or later. The pycycle-mcp server wraps cycle lifecycle, variable inspection and updates, execution, sweeps and total-derivative evaluation, with Pydantic checks and structured error payloads. Documentation

What NASA offers

The OpenMDAO site and documentation cover the framework, its components, solvers and derivatives, but no MCP server or AI assistant as of this check. The pyCycle server comes from a university group that also publishes MCP servers for SU2, TiGL geometry and mission analysis as part of a shared CPACS aircraft-analysis pipeline.

Checked 4 Nov 2026. Sources: OpenMDAO documentation, cmudrc/pycycle-mcp README

What you can do with OpenMDAO and AI

  • Create a pyCycle turbofan session, inspect its variables and change design inputs such as bypass ratio or pressure ratio.
  • Run the cycle and a parameter sweep, then read net thrust, TSFC, OPR, BPR and fuel flow.
  • Evaluate total derivatives of cycle outputs for a gradient-based study.
  • Size an engine to a target design thrust inside a CPACS aircraft pipeline alongside SU2 and mission-analysis servers.
  • Check structured error payloads when an input is out of range, rather than reading a Python traceback.

Taken from the listings' own documentation. Check each listing for exactly what it supports.

Related tools

OpenVSPNASA3 serversGMATNASA2 serversSU2Open source1 serverMATLABMathWorks7 servers · 2 skills

Questions

Is there an OpenMDAO MCP server?

Not for OpenMDAO in general. pycycle-mcp covers pyCycle engine-cycle models, which are built on OpenMDAO; for other models the agent writes and runs OpenMDAO Python directly.

Who maintains pycycle-mcp?

The Design Research Collective at Carnegie Mellon University, under the MIT licence. It is one of several servers (SU2, TiGL, mission segments, Aviary) that share a CPACS aircraft file in their pipeline.

Does it need a full pyCycle install?

Real cycle runs do. The repository's examples and tests are deterministic stand-ins, so development and CI work without the full pyCycle and OpenMDAO runtime. It needs Python 3.12 or later.

What does pycycle-mcp read and write in CPACS?

Its adapter reads engine data from the vehicles/engines branch and the aerodynamic drag coefficient from the aero results, runs a turbofan cycle, and writes design and off-design net thrust, TSFC, OPR, BPR and fuel flow to an mcpResults node under each engine.

Can Claude build other OpenMDAO models?

Not through a dedicated server. OpenMDAO models are Python, so an agent with a general code-execution tool can write components, groups and drivers, but there is no MCP server that validates or runs arbitrary OpenMDAO problems.

AI-generated models, drawings and results must be checked by a qualified engineer before use.

Sources (checked 4 Nov 2026): OpenMDAO documentation · OpenMDAO/pyCycle repository · cmudrc/pycycle-mcp README · OpenMDAO product page

OpenMDAO is a trademark of NASA. Tresca is independent and not affiliated with or endorsed by NASA.