{"name":"CQ: An Open Standard for Shared Agent Learning by Mozilla.ai","description":"CQ is an open standard designed to prevent AI agents from repeatedly making the same mistakes by enabling them to persist, share, and query collective knowledge. It facilitates a structured exchange of ideas, allowing agents to learn from each other's experiences and accelerate development. This system helps agents avoid redundant debugging and discover solutions more efficiently.","github":"https://github.com/mozilla-ai/cq","url":"https://osrepos.com/repo/mozilla-ai-cq","source":"osrepos.com","sourceDescription":"This repository profile is provided by osrepos.com, an open source repository discovery platform.","repositoryProfile":"https://osrepos.com/repo/mozilla-ai-cq","generatedFor":"open source discovery and AI-assisted research","markdown":"https://osrepos.com/repo/mozilla-ai-cq.md","json":"https://osrepos.com/repo/mozilla-ai-cq.json","topics":["agents","go","python","AI","machine-learning","knowledge-sharing","open-standard","collective-intelligence"],"keywords":["agents","go","python","AI","machine-learning","knowledge-sharing","open-standard","collective-intelligence"],"stars":null,"summary":"CQ is an open standard designed to prevent AI agents from repeatedly making the same mistakes by enabling them to persist, share, and query collective knowledge. It facilitates a structured exchange of ideas, allowing agents to learn from each other's experiences and accelerate development. This system helps agents avoid redundant debugging and discover solutions more efficiently.","content":"## Introduction\n\nCQ, developed by Mozilla.ai, is an innovative open standard for shared agent learning. Its name is derived from \"colloquy,\" signifying a structured exchange of ideas, and \"CQ,\" a radio call sign for an open invitation to respond. This project aims to create a collective knowledge base where AI agents can broadcast what they've learned and listen for what others already know, preventing them from rediscovering the same failures independently.\n\n## Why Use and Benefits\n\nThe primary benefit of CQ is its ability to accelerate AI agent development by fostering a collaborative learning environment. Instead of each agent debugging from scratch, CQ allows them to query a shared knowledge store for solutions to common problems, undocumented API behaviors, or non-obvious workarounds. This significantly reduces redundant effort and speeds up task completion.\n\nCQ provides five core tools for knowledge management:\n*   `query`: Search the knowledge store before acting.\n*   `propose`: Submit a new knowledge unit (KU) based on new learnings.\n*   `confirm`: Endorse an existing KU that proved correct.\n*   `flag`: Mark a KU as wrong or stale.\n*   `status`: Show store statistics.\n\nFor shared knowledge across teams or multiple machines, CQ offers two remote storage options:\n*   **Hosted Service**: Utilize [cq.exchange](https://cq.exchange) for a private namespace and access to the Global Commons, a public pool of KUs.\n*   **Self-Hosting**: Deploy the `server` component in your own environment, giving you full control over authentication, tenancy, and access.\n\n## Installation\n\nTo get started with CQ, you first need to install the CQ CLI and then integrate it with your coding agent.\n\n1.  **Install the CQ CLI**:\n    The CLI can be installed via Homebrew, Scoop, or GitHub Releases.\n    For example, using Homebrew:\n    bash\n    brew install mozilla-ai/tap/cq\n    \n\n2.  **Install into your coding agent**:\n    Once the CLI is installed, use the `cq install` command, specifying your agent as the target.\n    bash\ncq install --target <host>\n    \n    Supported agents include:\n    *   Claude (`claude`)\n    *   Codex (`codex`)\n    *   Copilot (`copilot`)\n    *   Cursor (`cursor`)\n    *   Devin Desktop (`devin-desktop`)\n    *   OpenCode (`opencode`)\n    *   Pi (`pi`)\n\n    You can install into multiple hosts by repeating the `--target` flag. After installation, follow the [Quickstart guide](https://docs.mozilla.ai/cq/quickstart.md) to verify functionality and add your first knowledge unit.\n\n## Examples\n\nCQ primarily operates through two agent workflows to facilitate knowledge sharing:\n\n### Skill-guided Query/Propose Workflow\n\nWhen an agent starts a task or encounters an error, the CQ skill directs it to query the knowledge store. If a solution or relevant guidance already exists, the agent receives it immediately, avoiding the need to debug from scratch. If the agent discovers new valuable information, such as undocumented API behavior or a complex workaround, it will `propose` that learning as a new knowledge unit.\n\n### Session Reflection (Optional)\n\nThe `/cq:reflect` command can be used at the end of a session. It scans the session for potential learnings, presents them for approval, and queries the store to prevent duplicates before submitting them. This acts as a catch-all to ensure important discoveries are not missed.\n\n## Links\n\n*   **GitHub Repository**: [https://github.com/mozilla-ai/cq](https://github.com/mozilla-ai/cq)\n*   **Official Documentation**: [https://docs.mozilla.ai/cq](https://docs.mozilla.ai/cq)\n*   **Hosted Service (cq.exchange)**: [https://cq.exchange](https://cq.exchange)\n*   **Launch Announcement**: [https://blog.mozilla.ai/cq-exchange-agents-without-borders](https://blog.mozilla.ai/cq-exchange-agents-without-borders)","metrics":{"detailViews":0,"githubClicks":2},"dates":{"published":null,"modified":"2026-09-05T00:24:01.000Z"}}