Mu…, name partly hiddenMu

Early plans. Everything on this page is subject to change while the design settles.

What it is

An AI agent factory for automated software development. Instead of one coding agent you steer step by step, it runs work through a production line: a task goes in, specialised agents plan it, build it, test it and review it, and a pull request comes out for a person to approve.

How it works

Issue → Plan → Build → Test → Review → Pull request

• Stations, not one agent: each step has its own agent with its own instructions, tools and permissions, so a reviewer never edits code and a builder never merges.
• Gates between stations: work moves on only when it passes checks: the plan is approved, the build compiles, the tests pass, the review is clean.
• Many lines at once: each task runs in its own isolated workspace, so several can be built in parallel without stepping on each other.
• People at the checkpoints: a person approves the plan and the final pull request, and can stop or redirect a line at any station.

Why I’m building it

Coding agents are good at single tasks but still need someone to plan, check and push every step along. A factory makes that the system’s job: the steps, the checks and the hand-offs are defined once and run the same way every time.

What you get

• A repeatable pipeline from issue to reviewed pull request.
• Defined roles for planning, building, testing and reviewing, each with only the access it needs.
• Visible progress: where every task is on the line, and why it stopped.
• Quality gates that use your existing tests and CI, not the agent’s opinion of its own work.

Status

Planned. Design first; the stations, gates and tooling above are expected to change.