dagoSte/squadra

What it solves

When you have several AI coding agents — like Claude Code, Codex, and Cursor — working on the same project, it quickly becomes messy: who owns which task, what has been verified, and who still needs to make a decision. Squadra fixes that by giving all agent windows a shared registry, a clear workflow, and independent reviews. Every change is checked by a different window than the one that wrote it, and every step leaves proof.

How it works

You install Squadra as a skill/plugin for your agents. A coordinator window (typically Claude) assigns tasks to builder windows, each working in its own checkout. Builders prove their work by running agreed checks on the exact files, then deliver to a reviewer window that has not touched the code. The reviewer records a verdict, and the coordinator checks and closes the task. All events — assignments, messages, checks, decisions — go into one shared registry on your computer. A local dashboard shows live progress, evidence, and when your input is needed. Squadra also tracks token usage per window, model, and task, and suggests which window should take the next task.

Who it’s for

Developers and teams who run multiple coding-agent sessions on the same project and want to keep control and quality. It is for people who value independent review, want proof tied to exact code, and prefer everything to run locally with no account and no cloud dependency.

Highlights

• One shared registry for every agent window — assignments, proofs, and decisions stay on your machine, not in chat memory.
• Independent review enforced: whoever writes a change does not review it.
• Proof is bound to the exact code; if the code changes after proof, the proof no longer counts.
• A task starts only after the recipient confirms receipt.
• When a decision is yours, you get options, a recommendation, and the exact words to answer.
• A plain-language dashboard (English or Italian) to follow everything live.
• Works with Claude Code, Codex, and Cursor, plus community plugins.
• Tracks token usage across windows and models with advice on which should handle the next task.