Honcho

Honcho is an open source memory library and managed service for building stateful AI agents. It enables agents to build and maintain state about any entity—users, agents, groups, ideas, and more—using a continual learning system that understands entities as they change over time. Honcho works with any model, framework, or architecture, and exposes a rich reasoning pipeline that asynchronously derives insights from interactions to personalize agent behavior.

• Peer Paradigm: Both users and AI agents are represented as "peers," enabling multi-participant sessions with mixed human and AI agents and flexible identity management.
• Memory & Storage Primitives: Workspaces, Peers, Sessions, Messages, Collections, and Documents provide a structured hierarchy for storing and retrieving all application and conversation state.
• Reasoning Pipeline: Asynchronously derives facts and representations about peers from messages, building comprehensive psychological models that inform agent responses.
• Chat (Dialectic) API: A natural language endpoint (/peers/{peer_id}/chat) lets developers query Honcho as an oracle about any peer—hydrating prompts, getting personalized insights, or requesting 2nd opinions.
• Context Endpoint: Returns a token-limited combination of messages, conclusions, and summaries to keep long-running LLM conversations going indefinitely.
• Hybrid Search: Query messages at the Workspace, Session, or Peer level using hybrid search with advanced filters.
• Representations: Low-latency static documents summarizing peer insights for quick prompt augmentation without waiting for an LLM response.
• Multi-Provider LLM Support: Configurable backends for Google Gemini, Anthropic, and OpenAI for different reasoning tasks.
• Self-Hostable: Full local development support via Docker Compose, Fly.io deployment, and a FastAPI server with PostgreSQL/pgvector.
• Python & TypeScript SDKs: Install via pip install honcho-ai or npm install @honcho-ai/sdk and get started with a few lines of code.