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.

What it solves

It prevents AI agents from repeating the same mistakes by providing a shared, structured knowledge store. Instead of every agent debugging a problem from scratch, they can access and contribute to a collective memory of solutions, undocumented API behaviors, and non-obvious workarounds.

How it works

cq implements a system where agents use a set of tools to interact with a knowledge store:

  • Querying: Before acting or retrying after an error, agents search the store for existing solutions.
  • Proposing: When an agent discovers a new solution or workaround, it proposes a "knowledge unit" to the store.
  • Reflection: Agents can scan an entire session to identify and submit learnings that were missed during the active workflow.

Technically, it operates via a local MCP server (written in Go) that manages a local SQLite database, which can optionally sync with a remote FastAPI server for team-wide or global knowledge sharing.

Who it’s for

Developers using AI coding assistants (such as Claude, Cursor, Copilot, or Devin) who want their agents to learn from previous experiences and share that knowledge across different sessions or team members.

Highlights

  • Multi-Agent Support: Compatible with a wide range of coding agents including Claude, Cursor, and Copilot.
  • Hybrid Storage: Supports local-only storage, a hosted service (cq.exchange), or self-hosted server deployments.
  • MCP Integration: Uses the Model Context Protocol (MCP) to provide tools for querying, proposing, confirming, and flagging knowledge.
  • Global Commons: Provides access to a shared public pool of knowledge units seeded by Mozilla.ai.

Related

  • Project
    TencentCloud/TencentDB-Agent-Memory

    TencentDB Agent Memory is an open‑source memory hub for LLM agents. It captures conversations, extracts reusable skills, builds a Wiki from docs, and indexes code into a CodeGraph. A central UI (Memory Hub) lets teams manage assets, set fine‑grained ACLs, and bind specific memory pieces to individual agents. Agents talk to a uniform proxy, so any framework (Claude Code, DeepSeek Harness, OpenClaw, etc.) can use the same memory without code changes. Benchmarks show a 59 % boost in persona retention. The repo provides one‑click Docker deployment, migration tools, and a roadmap for zero‑config cold starts and more integrations.

  • Project
    openclaw/agent-skills

    OpenClaw Agent Skills is a reusable library of workflow “skills” (e.g., autoreview, behavior validation, session viewer) for AI coding agents in OpenClaw projects. It provides a Python installer to symlink or copy these skills into a user’s agent‑skill directory, enabling consistent, version‑controlled automation across repositories. Includes validation scripts, MIT licensing, and support for Codex/Claude agents.

  • Project
    steipete/agent-scripts

    A centralized management system for sharing AI agent instructions, reusable skills, and helper scripts across local workspaces via symlink synchronization.

  • Project
    agentic-box/memora

    An MCP memory layer for AI agents that provides persistent collective memory through structured storage, semantic retrieval, and a knowledge graph.

  • Project
    mathomhaus/guild

    Guild is a Go‑based local server that gives AI coding agents persistent, searchable memory and a coordinated task board. It stores quests, lore entries, principles (oaths), and hand‑off briefs in SQLite, supports hybrid BM25 + vector search, provides atomic quest claims, and offers an agent‑friendly JSON CLI mode. Installable as a binary, Homebrew package, or Docker image, it lets agents across editors (Claude Code, Codex, Cursor) continue work seamlessly across sessions.