01 · AI Engineering Tools
Deterministic Engineering Tools for AI Assistants
Open Source Software Project · 2026
A TypeScript Model Context Protocol server that gives AI assistants five deterministic mechanical-engineering tools, replacing plausible guesses with validated, checkable calculations.
- 5
- MCP tools
- 26
- passing tests
- CI
- GitHub Actions
What
- An MCP server that exposes beam analysis, material properties, dimension-aware unit conversion, unit discovery, and experimental curve fitting to AI assistants.
- Designed for engineering workflows where a model's numeric answer must be deterministic and independently checkable.
How
- Built in TypeScript with the MCP SDK and Zod schema validation over a standard-input/output transport.
- Implemented ordinary-least-squares linear and power-law regression from scratch, including R-squared output and explicit invalid-input handling.
- Added 26 unit tests, GitHub Actions CI, an end-to-end demo, and setup instructions for Claude Desktop, Claude Code, and MCP Inspector.
Results
- Five callable tools return structured calculations instead of relying on model memory or arithmetic.
- The test suite validates known beam cases, material aliases, dimensional consistency, temperature conversion, and curve-fitting behavior.
- The open-source project includes a live demo GIF, CI status, installation steps, examples, and design notes.
Tools & methods
- TypeScript
- Model Context Protocol
- Zod
- Bun
- GitHub Actions
- Unit testing
Next project
Actuator Performance Analyzer