Talk to your logic analyzer.
Logic 2 now speaks MCP — connect it to Claude or Codex and control your analyzer with natural language.
Start captures, decode protocols, export data, and ask questions about your signals — all in natural language. No scripts, no boilerplate.
claude mcp add --transport http logic2 http://127.0.0.1:10530 Requires Logic 2 running locally with the MCP server enabled. See the docs for setup details.
Close the loop.
Your coding agent writes firmware. Now it can see what that firmware actually does on the wire.
An effective firmware agent needs to do everything a developer does: write code, flash the device, read UART logs, and verify what's happening on the bus. The tools to give your agent access to flashing, serial output, and device reset exist today — but electrical measurement has been the missing piece.
Logic 2's MCP server closes that gap. Once your agent can capture and decode real signals, the full loop is closed: write, flash, capture, verify, iterate — no human in the loop.
This is the first step in closing the LLM loop for electrical measurement and verification. See the docs to get started.
Tests that write themselves.
Your agent captures real signals, then writes and maintains the tests to verify them.
Hardware test coverage is hard. Writing tests against protocol traces and voltage levels is tedious, brittle, and the first thing that falls behind when deadlines hit.
With an agent connected to your Logic analyzer, the dynamic changes. It captures real signals, understands the protocols, and writes test assertions against what it actually sees — not rubber-stamp pass/fail, but assertions grounded in what the spec says the bus should look like.
When firmware changes, the agent re-captures, compares against the spec, and flags real regressions instead of blindly updating assertions to match new output.
See the docs to connect your Logic analyzer to Claude Code or Codex.
- Hand-write assertions against captured data
- Tests break when firmware changes — someone has to triage each one
- Coverage gaps where writing tests is too tedious
- Validation bottlenecked on one engineer who knows the protocol
- Agent captures real traces and writes test code against them
- Firmware changes? Agent re-captures and flags real regressions vs. expected changes
- Coverage across SPI, I2C, UART, and custom protocols
- Runs in CI — hardware regressions caught before merge