Why it mattersA developer who runs more than one AI coding agent can now track every task on one dashboard. The tool refuses to close a task when the writer and reviewer are the same window.
A developer running three AI coding windows cannot tell which one wrote which change. Squadra is a new open-source Python tool from the developer dagoSte. It sits above Claude Code, Codex and Cursor and gives each window its own tasks. The repository has 419 stars after 26 days on GitHub.
Squadra is MIT-licensed. It is written in Python 3.12 and uses the standard library only. It runs on macOS and other POSIX systems. The current version is 3.10.4. The dashboard is a local web page.
What the dashboard shows
The dashboard has pages for tasks, communications, decisions, usage, balance, runs, outputs, timeline and history. The usage page tracks tokens by window and by model. The balance page suggests how to load the next task. The tool refreshes the dashboard every two seconds. The README reports a speed change in version 3.10.2. Each page update now takes 0.17 seconds. In version 3.10.1 it took 4.7 to 5.3 seconds.
Source-bound proofs
The author calls one feature source-bound proofs. The README states: "Proof bound to the exact code. If the code changes after the proof, proof and review no longer count." A test run and a review are tied to one commit. Any later change marks them stale and the task cannot close until a new proof and review arrive.
One window cannot review itself
The coordinator assigns the reviewer. Version 3.10.4 added a rule that the review has to come from a different model. "Codex writes and Claude reads, Claude writes and Codex reads." A same-model review is refused unless the user documents a reason with the flag --same-client-reason.
The README gives one example. A key rotation service is written by one window, which runs the test suite and freezes the change. A second window reads the candidate. The coordinator then closes the task.
Where this fits
A developer with one coding agent does not need Squadra. The audience is a working developer who already runs two or three agents at once and loses track of which window owns which change. The dashboard puts every task, every token count and every decision in one place. The cross-model review rule turns the common failure mode, one agent approving its own code, into a hard refusal.
Squadra is a tool for a developer who has already decided to run several agents. Reviewing an agent on its own output is the first thing to go when a project moves fast, and the tool now makes that gap visible instead of letting it close silently.
Source
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
