Automation · AI · API

Documentation

How to automate a bench in BenchMesh - by hand, by describing it to a model of your choice, or from an AI assistant that talks to the app - and how to run it on a Mac.

01Automation with flowsRead instruments, decide, set, wait, loop, sweep and record - flows are JSON documents the app checks against your bench and runs with confirmation of every change.Read →
02Build flows by describing themDescribe an automation in plain words and an AI model of your choice assembles the flow. Bring your own key: Anthropic, OpenAI, OpenRouter or a local model.Read →
03Connect AI assistants (MCP)Let Claude Code, Claude Desktop, Cursor and other MCP clients read your bench, write flows and - when you allow it - run them.Read →
04Run BenchMesh on macOSWhich .dmg to download, the one-time step that opens an app macOS calls “damaged” or “could not verify”, and instruments over USB-TMC and USB-serial on a Mac.Read →

HTTP APIs

The app’s local API (instruments, recordings, automation) is described by the OpenAPI document it serves at http://127.0.0.1:57666/docs; the automation endpoints are listed in Automation → REST API. This website’s read-only instrument database and community status: Public API.