mu

Coding agent built on pi that hands routine decisions to a small judge model, so the main model can focus on the code.

Agent by qybaihe

IndexedMIT★ 544

Write a review

Get started

Setup steps are in the repository. Website · Repository

About

mu is a coding agent, offered as a command-line tool and a desktop app, that sends its routine decisions, such as what enters the context, whether a command is safe and whether work is done, to a small judge model at about 40 decision points per session. The default judge is Jev, hosted by TypeSafe, and any judge can be swapped in, including the mude 1.0 model that runs on your Mac or any LLM. Every verdict goes to a local ledger that can be audited, and in the default Judge mode the judge lets through what the task clearly needs and asks you about the rest.

Use cases: coding, code-generation

Reviews

Write a review

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Fact sheet

Jobs: codingREADME
Autonomy: Supervised: runs multi-step tasks, asks at checkpointsREADME
Where you use it: Terminal, Desktop appREADME
Hosting: Hosted or self-hostedREADME
Models: jev, mude-1.0README
Interfaces: mcpREADME
Open source: YesREADME

Facts extracted by Claude Haiku 5.5 from the linked sources on Oct 24, 2026. Wrong fact? Tell us.

Signals

Stars: 544
Forks: 63
Open issues: 0
Last commit: Oct 24, 2026 (4 days ago)
Last release: Oct 24, 2026 (4 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License: MIT
Repository: github.com/qybaihe/mu
Website: usemu.dev
First seen: Oct 24, 2026
Last checked: Oct 24, 2026
Source: README github-graphql