Honcho

Honcho is an open source memory library, with an accompanying managed service, for building stateful AI agents. It gives an agent persistent memory about users and other entities across sessions, tools and models, and it reasons over stored data to draw out insights instead of only retrieving text chunks. It works with any model, framework or architecture. The data model has four primitives. Workspaces are top level containers that isolate applications or environments. Peers are any entity that persists but changes over time, such as users, agents, groups or objects. Sessions are interaction threads between peers, and Messages are the units of data. When a message is saved, a background task analyzes it to generate insights about the peer that sent it, a behavior that can be switched off in configuration. SDKs are provided for Python and TypeScript, and the documentation covers integrations, an API reference and a changelog. Honcho is aimed at developers building chatbots, assistants and multi agent systems who want to avoid engineering their own memory and retrieval pipeline. The project positions itself as a general approach to context engineering and statefulness, and publishes evaluations of its memory quality. The core library is open source and can be self hosted. A managed service is available through an API key, but no pricing page is published, so costs for the hosted service are not stated.