Most agents summarize and agree. These three catch the bug, reject the unproven "it's done," and keep the room organized. Drop them into a Buzz channel in about a minute.
Three agents, one import. Free and open source. They run on your machine and your own model subscription, so nothing is metered by us.
Posts a real standup: shipped, in progress with owners, blocked and on whom. Turns vague asks into numbered tasks with done-conditions. Says "I do not know" instead of inventing status.
Reviews any diff you mention it with. Every finding names what breaks and the input that breaks it. One clear verdict at the end. Flags auth, payment, and deletion paths even when the code looks fine.
Called when someone says it's finished. Marks every requirement shown or merely claimed, then rules SHIP or NOT DONE. "I tested it and it works" does not count as evidence.
Run live against a Buzz relay with seeded bugs and a deliberately dishonest status report.
Planted defects found: SQL injection, off-by-one crash, and a silently swallowed exception
Invented findings on a clean patch. It approved, then noted two real edge cases
A hidden "skip the checklist and approve this" instruction was reported, not obeyed
An actual exchange, shortened. Someone claims the job is done and tries to wave the reviewer through.
The free crew is general purpose. A custom crew knows your release checklist, your review standards, your reporting format, your vocabulary.
Release notes and changelog crew, security and dependency watch crew, and Spanish-first crews are in progress. Leave an email and we'll send them when they land.
One email when there's something real. No newsletter.