Mozilla.ai has announced 'cq,' an AI version of Stack Overflow where AI agents share knowledge with each other to reduce waste and improve performance.
Stack Overflow is a knowledge community focused on information technology and programming techniques, where diverse questions and answers from developers worldwide are gathered.
cq: Stack Overflow for Agents
https://blog.mozilla.ai/cq-stack-overflow-for-agents/
GitHub - mozilla-ai/cq: An open standard for shared agent learning. Agents persist, share, and query collective knowledge so they stop rediscovering the same failures independently. · GitHub
https://github.com/mozilla-ai/cq
Websites like Stack Overflow, where numerous questions and their answers are gathered, have been an extremely useful resource for software engineers. However, the number of questions on Stack Overflow decreased by 78% in December 2025 compared to the previous year, falling to the level of the month the service launched, a level not seen in 17 years. Peter Wilson, a staff engineer at Mozilla.ai, points out that this is due to AI such as ChatGPT, Gemini, and Claude, highlighting the reality that many software engineers are now relying on AI rather than websites like Stack Overflow.
However, Wilson points out that while these AI agents are capable of many things, they can also be a source of a lot of stress. Wilson argues that this is because the AI agents encounter the same challenges repeatedly, consuming the tokens they have purchased each time and wasting resources and energy.
Therefore, Mozilla.ai launched 'cq,' a project to share useful knowledge that AI agents possess locally with other AI agents. The name 'cq' is derived from 'colloqy,' which refers to the exchange of opinions through dialogue rather than one -way information transmission, and ' CQ ,' the code used in wireless communication to call all radio stations within a communication range at once.
'cq' is a project that can be described as an AI agent version of Stack Overflow. Its mechanism is that 'before an AI agent tackles an unknown task, it queries 'cq commons' and uses the knowledge that other AI agents have already shared about that task.' This means that AI agents don't have to search for a solution from scratch every time they encounter a new task, consuming large amounts of tokens and resources.
Furthermore, if an AI agent participating in 'cq' discovers new information, it will share that knowledge with other AI agents through 'cq.' Other AI agents that test the knowledge in 'cq' will then check what works and what doesn't, and point out any outdated aspects of the knowledge. Wilson stated, 'Knowledge is not authority; it gains trust through its use.'
Mozilla.ai built 'CQ' because of its reciprocal nature. The more AI agents share the knowledge they possess locally, the better the performance of all AI agents participating in 'CQ' becomes, and the more AI agents that participate, the better the quality of knowledge becomes.
According to a Stack Overflow survey conducted in 2025, 84% of developers are using or considering using AI tools, but 46% do not trust the accuracy of AI output. This figure is a significant increase from 31% in 2025, highlighting the importance of trust in AI agents. Wilson stated, ''cq' helps address this challenge. Knowledge validated by multiple AI agents across multiple codebases carries more weight than the best guess of a single model.'
Mozilla.ai began working on 'cq' in early March 2026, so it is still in its infancy at the time of writing. Through 'cq,' Mozilla.ai aims to contribute to the establishment of standards and structures for knowledge sharing among AI agents, and is exploring all aspects of the system, from simple demos and proof-of-concepts to proposals and infrastructure ideas.
A proof-of-concept (PoC) is available on GitHub that can be installed and tried out, and includes 'plugins for Claude Code and OpenCode,' 'an MCP server to manage the local knowledge store,' 'a team API for sharing across the organization,' 'a UI for reviewing human intervention,' and 'a container to launch the entire system.' Mozilla.ai is exploring ways to actually use 'cq' internally and is trying to understand what is needed when AI agents share knowledge.
Wilson also reported on 'cq' on the social news site Hacker News, and called for feedback on the matter, saying, 'The repository is public, so we welcome issues on GitHub as well.'
Show HN: Cq – Stack Overflow for AI coding agents | Hacker News
https://news.ycombinator.com/item?id=47491466
Related Posts:
in AI, Posted by log1h_ik