Cq: Mozilla AI Builds Stack Overflow for AI Coding Agents
Mozilla AI has launched Cq, an open-source knowledge-sharing platform where AI coding agents can learn from each other's experiences — essentially a Stack Overflow for agents.
The name comes from "colloquy" (structured dialogue) and the radio term "CQ" (a general call for response). Before an agent tackles an unfamiliar task — API integrations, CI/CD configs, framework quirks — it queries the Cq commons. If another agent has already discovered, say, that Stripe returns 200 with an error body for rate-limited requests, the new agent knows that before writing a single line of code. When agents discover novel solutions, they propose that knowledge back for community validation.
The system includes trust mechanisms: knowledge confirmed by multiple agents across multiple codebases carries more weight than a single model's best guess. The proof-of-concept ships with plugins for Claude Code and OpenCode, an MCP server managing local knowledge stores, a Team API for organizational sharing, and a UI for human-in-the-loop review.
Cq addresses a critical inefficiency in the current agent ecosystem: every agent independently solves the same problems, burning tokens and compute each time. Andrew Ng recently posed the same question — whether there should be a Stack Overflow for AI coding agents — validating the category.
GitHub: https://github.com/mozilla-ai/cq
Blog post: https://blog.mozilla.ai/cq-stack-overflow-for-agents/
← Back to all articles
The name comes from "colloquy" (structured dialogue) and the radio term "CQ" (a general call for response). Before an agent tackles an unfamiliar task — API integrations, CI/CD configs, framework quirks — it queries the Cq commons. If another agent has already discovered, say, that Stripe returns 200 with an error body for rate-limited requests, the new agent knows that before writing a single line of code. When agents discover novel solutions, they propose that knowledge back for community validation.
The system includes trust mechanisms: knowledge confirmed by multiple agents across multiple codebases carries more weight than a single model's best guess. The proof-of-concept ships with plugins for Claude Code and OpenCode, an MCP server managing local knowledge stores, a Team API for organizational sharing, and a UI for human-in-the-loop review.
Cq addresses a critical inefficiency in the current agent ecosystem: every agent independently solves the same problems, burning tokens and compute each time. Andrew Ng recently posed the same question — whether there should be a Stack Overflow for AI coding agents — validating the category.
GitHub: https://github.com/mozilla-ai/cq
Blog post: https://blog.mozilla.ai/cq-stack-overflow-for-agents/
Comments