honcho

Honcho addresses the challenge of building stateful AI agents that maintain and reason over evolving knowledge about people, agents, groups, projects, and ideas. Unlike simpler vector store-based memories that rely on chunk matching, Honcho applies reasoning-first memory to extract conclusions from conversations and events, supporting multi-peer perspectives and peer-centric models. This makes it suitable for teams developing AI systems that require nuanced, context-rich memory over time. The project is moderately mature with nearly 7,000 stars and recent commits, but it lacks significant growth momentum, suggesting a niche or early-stage adoption. The core is implemented as a FastAPI server with Python and TypeScript SDKs, enabling integration with tools like Claude Code and Cursor. It is not ideal for teams seeking a lightweight or purely embedding-based memory store, where alternatives like LangChain or Weaviate may be more appropriate due to their broader ecosystem and simpler setups.

Use Honcho when building AI agents that need to maintain long-term, evolving memory with reasoning capabilities about multiple interacting entities. If your use case is limited to embedding similarity search or simpler memory without complex peer models, consider LangChain or Weaviate instead for their more established ecosystems and simpler integration.

Honcho typically fits into systems requiring persistent, reasoning-enabled memory for AI agents such as coding assistants (e.g., Claude Code, Cursor), collaborative bots, or multi-agent frameworks that track relationships and context over time. It is well suited for products that need to maintain stateful, multi-peer knowledge rather than one-off retrieval or static document search.

Latest commit: 2026-09-02. Activity data is based on daily RepoPi snapshots of the GitHub repository.