Hi everyone. 
I’d like to share an open-source project I’ve been working on and gather
feedback from the QSpice community.
What is it?
qspice-mcp is a Model Context Protocol (MCP) server that connects
QSpice to AI assistants (Claude, Cursor, VS Code Copilot, etc.). It lets
the AI inspect and edit .qsch schematics, run simulations, read waveforms,
and export results, all through natural language, using stable JSON tools.
GitHub: Why I built it
While working on QSpice projects, I noticed a lot of repetitive tasks
(editing components, sweeping parameters, parsing .raw files, comparing
runs). MCP makes it possible to delegate these to an AI assistant while
keeping QSpice as the simulation engine. The goal is to help engineers,
students, and makers iterate faster, not to replace the simulator or
the engineer’s judgment.
Key features
Inspect & edit .qschschematics programmatically
Run QSpice simulations from the AI assistant
Read and analyze waveform / .rawoutput
Built-in workflow examples (e.g., buck converter, C-Block DLL devices)
Stable JSON tool contracts, predictable for AI agents
MIT licensed, with documented threat model and security policy
Requirements
- Windows
- Python ≥ 3.11 (tested on 3.11 and 3.12)
- QSpice (
QSPICE64.exe)
Setup is a few commands; full instructions and a user guide are in the repo.
What I’m looking for
This is an early community project and I’d really appreciate your input:
- Bug reports / issues — please open a GitHub Issue with reproduction steps
- Feature requests — which QSpice workflows would you most like to automate?
- Feedback on tool design — are the JSON tool contracts clear and useful?
- Example schematics — if you have circuits that are good stress tests, I’d love to try them
- General thoughts — does this fit how you actually use QSpice?
Thanks for taking a look; looking forward to your feedback. 
H.Abedini
Hossein Abedini - Freiburg, Baden-Württemberg, Germany | Professional Profile | LinkedIn