cq
An open standard and CLI for shared agent learning, letting AI coding agents query and contribute a structured store of prior fixes and workarounds so they stop rediscovering each other's mistakes.
cq is an open standard for shared agent learning: structured knowledge that keeps AI agents from repeating each other's mistakes. The name draws on 'colloquy,' a structured exchange where understanding emerges through dialogue, and 'CQ,' a radio call sign meaning any station, respond, both describing the same idea of agents broadcasting what they have learned and listening for what others already know.
In practice, when an agent starts a task or hits an error, a cq skill directs it to query the shared knowledge store before retrying, so a problem another agent already solved does not have to be debugged from scratch; if the agent discovers something worth sharing, such as undocumented API behavior or a workaround for a known issue, it proposes that learning as a new knowledge unit. A /cq:reflect command scans a finished session for anything the agent may have missed proposing.
The CLI installs through Homebrew, Scoop, or GitHub Releases and then installs itself into a coding agent such as Claude, Codex, Copilot, Cursor, Devin Desktop, OpenCode, or Pi; the project is at an early 0.x stage with expected breaking changes.