Honcho logo

Honcho

Connects coding agents to Honcho’s persistent memory for storing conversations, querying knowledge, search, and context retrieval.

Install Honcho

Pick your client, then add it in one click or paste the config.

STDIOPython
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "honcho": {
      "command": "uvx",
      "args": [
        "honcho-ai"
      ],
      "env": {
        "HONCHO_API_KEY": "YOUR_VALUE_HERE"
      }
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

Values are written into the snippet above so you can paste it as is. They never leave your browser.

Install not yet sandbox-tested— we haven't run this command through our automated sandbox yet. Not a red flag; we're working through the catalog.

Overview

Honcho MCP server connects coding agents and other MCP clients to Honcho’s memory infrastructure for stateful interactions. It stores messages and events, processes them asynchronously, and exposes peer knowledge, session context, search results, representations, and workspace-level answers. Use it when an agent needs continuity across conversations or must retrieve context without placing an entire history in every prompt. Honcho can use the managed service at api.honcho.dev, a local stack, or a self-hosted FastAPI deployment.

Use cases

  • Give coding agents persistent memory across sessions
  • Store user conversations and retrieve relevant context
  • Search workspace knowledge with hybrid retrieval
  • Ask natural-language questions about a peer or workspace

Key features

  • Persistent workspace, peer, and session memory
  • Asynchronous reasoning over messages and events
  • Hybrid BM25 and vector search
  • Prompt-ready session context
  • Peer and session representations
  • Managed, local, or self-hosted deployment

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked Today

Recent health check succeeded.

40/40 checks healthy over the last 57d

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 6d ago via OSV.dev · honcho-ai (PyPI)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars8kStargazers on the source repository.
Last commitTodayMost recent push to the default branch.
Directory activity7 viewsConfig copies, upvotes, and views on AllMCPs.

How Honcho works

What Honcho MCP server does

Honcho MCP server gives MCP-compatible agents access to Honcho’s memory model. Workspaces contain peers, peers participate in sessions, and messages are stored on those sessions. Honcho then builds representations of people, agents, groups, projects, and ideas as conversations and events accumulate.

The memory workflow has four stages: record interactions, process them in the background, query the resulting knowledge, and add the returned material to a model prompt. This supports agents that need to remember user preferences, prior work, or evolving relationships rather than treating each request as isolated.

How it works

Messages, documents, and tool traces can be added to a session. Honcho processes queued data asynchronously and updates peer representations. An agent can then request a natural-language answer about a peer, retrieve context for a session, search stored information, or read a lower-latency representation.

Available memory operations include peer, session, and workspace-level queries. Search combines BM25 and vector retrieval. Session context can be formatted for downstream model APIs, including OpenAI and Anthropic message formats, before being injected into an agent call.

Because reasoning runs in the background, recently added messages may not immediately appear in chat or representation results. The representation endpoint is intended for lower-latency reads when static, already-processed information is sufficient.

Setup and configuration

Honcho offers a managed deployment at api.honcho.dev. Create an account through app.honcho.dev and use an API key with the workspace configured for the agent. The Python and TypeScript SDK examples identify the primary settings as a workspace ID and HONCHO_API_KEY.

For local use, install the Honcho CLI and start the local stack with honcho start --setup; clients can then target http://localhost:8000. Self-hosting is also available by running the FastAPI service from source or with Docker Compose. Set HONCHO_URL when the service is not using the managed default.

MCP clients named for Honcho include Claude Code and Cursor, alongside other coding-agent integrations and MCP-compatible clients. The same Honcho memory service can be used by these agents, so stored information can support more than one client.

Tools and capabilities

  • Store conversations, events, documents, and tool traces as session messages.
  • Ask for a natural-language answer about a peer or across a workspace.
  • Retrieve prompt-ready session context with a token limit and optional summary.
  • Search peers, sessions, or the workspace using hybrid BM25 and vector retrieval.
  • Read peer or session representations for lower-latency access.
  • Upload documents into a session.
  • Check the status of background processing queues.
  • Model relationships between peers when multi-peer perspective is configured.

The underlying service also provides Python and TypeScript SDKs, which can be used when an application needs direct integration outside an MCP client.

Limitations and notes

Background reasoning is asynchronous, so newly stored messages may require processing before chat and representation queries reflect them. For immediate reads, choose the representation endpoint where its static result is appropriate. Honcho’s memory is organized around workspaces, peers, sessions, and messages; applications should choose stable identifiers for those entities to preserve continuity.

The system can be managed, run locally, or self-hosted. A managed setup requires an API key, while a local or self-hosted deployment requires the client to point at the selected Honcho URL.

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Honcho

Honcho is an MCP server that connects coding agents and other MCP clients to Honcho’s memory infrastructure. Its main tools cover storing session messages, querying peers or workspaces, retrieving context, searching memory, reading representations, uploading documents, and checking background processing status.

Use Honcho through its managed service, a local stack started with `honcho start --setup`, or a self-hosted FastAPI deployment. Configure the MCP client to use the selected Honcho service and provide the workspace and API-key settings required by that deployment.

Yes, Cursor is listed as a Honcho coding-agent integration. Cursor can use Honcho’s shared memory service to support persistent context across agent interactions.

Honcho processes newly stored messages and events asynchronously, then updates peer representations that agents can query. Use chat, search, session context, or representations to retrieve the resulting knowledge.

Yes, Honcho can run locally or be self-hosted. The CLI can start a local stack with `honcho start --setup`, and clients can target `http://localhost:8000`; Docker Compose is also available for self-hosting.

Related MCP servers

Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewHoncho AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/honcho?style=directory)](https://allmcps.com/mcp/honcho)
HTML Embed
<a href="https://allmcps.com/mcp/honcho"><img src="https://allmcps.com/api/badge/honcho?style=directory" alt="Honcho on AllMCPs" /></a>