Awesome AI Agents › AI Agent Development

plastic-labs/honcho

⭐ 7151 Python added to this list on 2026-04-27 repository created 2023-09-10

Honcho is a robust memory infrastructure designed to enable the creation of stateful AI agents capable of understanding and adapting to evolving entities such as people, other agents, groups, projects, and ideas. It functions by storing messages and events, allowing Honcho to asynchronously reason in the background to build and update peer representations. This system supports querying for session context, peer representations, search results, or natural-language insights, which can then be injected into various LLM calls or agent frameworks. Key features include a 'reasoning-first' memory model that extracts conclusions from conversations rather than just matching content chunks, and a 'peer-centric' approach that tracks and models how different entities change over time. It also offers a 'multi-peer perspective,' modeling what one peer knows about another when configured. Honcho can be used as a managed service at api.honcho.dev or self-hosted as a FastAPI server. It provides SDKs for Python and TypeScript, making it accessible for developers to integrate persistent memory into their agents and applications. The project aims to improve agent retention, foster trust, and help build data moats through superior memory management, as evidenced by its benchmarks and evaluations. It is split into several repositories, with this one hosting the core service logic. Honcho specifically focuses on enabling richer, more adaptive, and long-term interactions for AI agents.

https://github.com/plastic-labs/honcho

agentaichatbotcontextembeddingsllmmemorypersonalizationragstatefulfastapipythontypescriptmulti-agentsemantic memorylong-term memory