</>MCP Agents Market
Plugin

honcho

by plastic-labs6.9kPythonUpdated 2026-08-27

Memory library for building stateful agents

Claude CodeCodexCursorOpenCodeOpenClawHermesClaude DesktopWindsurf

Honcho is a memory infrastructure library that enables AI agents to maintain stateful, long-term memory about users, conversations, and context over time. It provides a reasoning-first memory system that extracts conclusions from interactions rather than just matching text chunks, modeling users, agents, groups, and projects as evolving entities. Developers can integrate Honcho via Claude Code plugins, MCP servers, or direct SDK usage (Python/TypeScript) to give their agents higher retention, personalized responses, and the ability to recall past interactions. The service runs as a managed API at api.honcho.dev or can be self-hosted locally via Docker.

Key Features

Reasoning-first memory that extracts conclusions and insights from conversations, not just keyword matching
Peer-centric model tracking users, AI agents, groups, and projects as first-class entities that change over time
Multi-participant sessions with configurable observation settings between peers
Chat endpoint for natural-language queries about what the system knows ("What learning styles does the user respond to best?")
Hybrid search (BM25 + vector) across workspaces, sessions, and peers with advanced filtering
Context endpoint that returns prompt-ready message bundles with conclusions and summaries up to a token limit
First-party integrations for Claude Code, Codex, Cursor, OpenCode, OpenClaw, Hermes, and any MCP client
Managed cloud service with $100 free credits or self-hostable via CLI (honcho start) or Docker Compose

Use Cases

  • 01Building coding assistants that remember user preferences, coding style, and project context across sessions
  • 02Creating tutoring agents that track learning progress and adapt to individual student needs over time
  • 03Developing multi-agent systems where agents maintain shared memory and understanding of each other
  • 04Enabling long-running conversations that exceed context windows by summarizing and retrieving relevant history
  • 05Personalizing AI responses based on accumulated knowledge about user behavior and preferences
  • 06Building data moats for AI products by storing rich user interaction history and insights

Related Plugins

View more

honcho — FAQ

What is Honcho?+

Honcho is a memory infrastructure service that gives AI agents persistent, reasoning-based memory about users, conversations, and context. It stores messages and events, reasons about them in the background to extract conclusions, then provides query APIs to retrieve insights, context, and search results that can be injected into any LLM call.

How do I install the Honcho plugin for Claude Code?+

Run '/plugin marketplace add plastic-labs/claude-honcho' then '/plugin install honcho@honcho' in Claude Code. Alternatively, for raw MCP integration in any client, run 'claude mcp add honcho --transport http --url https://mcp.honcho.dev --header "Authorization: Bearer hch-your-key-here"'.

Which AI clients work with Honcho?+

Honcho has first-party plugins for Claude Code, Codex, Cursor, OpenCode, OpenClaw, DeepSeek Harness, and Hermes. It also works with any MCP-compatible client (Claude Desktop, Cline, Windsurf, etc.) via its MCP server endpoint. For custom integrations, use the Python or TypeScript SDKs.

Do I need an API key to use Honcho?+

Yes. Sign up at app.honcho.dev to get an API key with $100 free credits on the managed service. Alternatively, run 'uv tool install honcho-cli && honcho start --setup' to run a local instance (requires Docker and an LLM provider API key like OpenAI, Anthropic, or Gemini).

Is Honcho free and open source?+

Honcho is open source under AGPL-3.0 and can be self-hosted for free. The managed cloud service at api.honcho.dev offers $100 in free credits for new organizations, with usage-based pricing thereafter.

How does Honcho differ from a vector database or RAG system?+

Honcho uses reasoning to extract conclusions and insights from interactions rather than just storing and retrieving text chunks. It models users, agents, and groups as evolving entities with representations, peer cards, and session summaries, enabling more sophisticated personalization than traditional RAG approaches.

How do I install honcho?+

Open the source repository on GitHub and follow its README. honcho is a plugin — MCP Agents Market links you directly to the official repo.

Is honcho free?+

honcho is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches