Honcho

Official

Reasoning memory for stateful agents.

How to install ↓ Compare GitHub source repository ↗

Category: Other #106 of 230
Stars: ★ 7.6k Very popular
Transport: Streamable HTTP
Runtime: Hosted (no local runtime)
Credentials: API key / credential required
License: AGPL-3.0
Last commit: 2 days ago

55FMRS · C

This official Honcho MCP server provides stateful-agent memory over streamable HTTP. The source documents the remote endpoint and authentication requirements, but does not list MCP tool identifiers, so the tool list is empty.

Strongest · Documentation 14/20 Weakest · Reliability 8/20

Reliability

8/20

Security and permissions

9/20

Maintenance

11/20

Documentation

14/20

Setup experience

13/20

Why each score

Reliability 8/20: The manifest declares a Streamable HTTP endpoint and a required Authorization header, but provides no tool list, parameter schemas, error semantics, or real CI/test evidence. The MCP initialization, tool-list handshake, and alignment between declarations and implementation therefore cannot be verified statically. The README also shows both 3.0.0 and 3.0.9 version information, so the score is conservative.
Security and permissions 9/20: Authorization is marked required and secret, supports OAuth tokens or Honcho API keys, and an optional workspace header provides a basic data-boundary mechanism. Deductions apply because the hosted endpoint receives and processes memory, messages, embeddings, and LLM-related data, while the source provides no evidence of tool-level least privilege, retention controls, complete data-flow disclosure, dangerous-operation confirmation, or MCP-level access controls. The self-hosting example also shows authentication disabled by default. No stated security red line is evidenced.
Maintenance 11/20: The repository is stated to be unarchived, the README presents server and SDK version information, and the AGPL-3.0 license is explicit; these support ongoing maintenance and clear licensing. Deductions apply because the supplied material contains no commit history, CI status, dependency-update policy, vulnerability-response channel, or explicit maintenance commitment. The issue count does not add points.
Documentation 14/20: The README covers managed and self-hosted installation, Python and TypeScript examples, authentication, architecture, asynchronous processing, configuration, integrations, and operational components. Deductions apply because MCP tool names, parameters, return values, error behavior, limits, costs, privacy/retention, and troubleshooting are not evidenced sufficiently, and the manifest 3.0.0 conflicts with the README's 3.0.9 server badge.
Setup experience 13/20: The source provides a hosted MCP URL, Authorization format, optional workspace header, and configuration examples for Claude Code and other MCP clients; it also documents Docker and local FastAPI/worker paths. Deductions apply because obtaining and configuring credentials remains manual, client coverage is limited, and self-hosting requires a database, environment variables, migrations, an API server, and a background worker, with no verifiable execution evidence.

Static review · not runListed 2026-08-14

Read the FMRS scoring method →

Fit and risk

What it can accessUses the network

Best for

Stateful agents that need cross-session memory
Applications requiring personalization and long-term context
Developers using MCP-compatible clients
Teams choosing between a managed or self-hosted memory service

Not for

One-off conversational applications that do not retain history
Integration reviews requiring a documented list of MCP tool names and parameters
Workloads where messages or documents must not be sent to an external memory service

Required permissions

Requires an Authorization request header to mcp.honcho.dev
Requires a valid OAuth access token or Honcho API key
May optionally require a default workspace ID
The service may receive and process session messages, events, documents, or tool traces

Risks and side effects

Content sent to Honcho may contain sensitive personal or application data
Background reasoning is asynchronous, so newly added messages may not immediately appear in chat or representation results
The managed service depends on network access, credentials, and Honcho availability
The AGPL-3.0 license may affect self-hosting and redistribution plans

Setup

Before you start

Runtime：Hosted (no local runtime)

Connect an MCP-compatible client to https://mcp.honcho.dev. The Authorization header is required and must use Bearer <token>; the token may be an OAuth access token or a Honcho API key from app.honcho.dev, whose keys start with hch-. Optionally set X-Honcho-Workspace-ID to provide a default workspace.

.mcp.json

{"mcpServers":{"honcho":{"url":"https://mcp.honcho.dev","headers":{"Authorization":"Bearer <token>"}}}}

Shown for Claude Code. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json

{
  "servers": {
    "honcho": {
      "url": "https://mcp.honcho.dev",
      "headers": {
        "Authorization": "Bearer <token>"
      }
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal

claude mcp add --transport http honcho https://mcp.honcho.dev --header 'Authorization: Bearer <token>'

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

Connect the client to https://mcp.honcho.dev with an Authorization: Bearer <token> header; on success the client should list Honcho's tools. Without a valid token the connection fails.

Troubleshooting

Verify that Authorization uses the Bearer <token> format
Confirm that the token is valid and that Honcho API keys start with hch-
Confirm that the MCP client supports streamable HTTP
Check whether X-Honcho-Workspace-ID points to the intended workspace
After writing new messages, allow time for background queue processing
For self-hosting, check the database, required LLM API keys, migrations, API server, and background deriver

Things to try

Once connected, you can ask your AI assistant things like:

Query my peer's representation in Honcho
Store this conversation in a Honcho session
Search workspace messages for what the user prefers
Generate a context summary for this session

Use cases

Give coding agents persistent memory

Store conversations, events, documents, and tool traces

Query long-term representations and insights about users or agents

Generate session context for injection into LLM prompts

Perform hybrid message search

Supported clients

Claude Code

CursorPartial support

ClinePartial support

WindsurfPartial support

OpenCodePartial support

OpenClawPartial support

HermesPartial support

Listed from the project's documentation, not tested by this site.

Overview

Honcho is memory infrastructure for building stateful agents. It stores messages, events, documents, or tool traces, reasons over them in the background, and builds evolving representations of people, agents, groups, projects, and ideas. Applications can query context, search results, representations, or natural-language insights. It is available as a managed service or as a self-hosted FastAPI server.

Similar servers

Hevy MCP 77 · B

Manage your Hevy workout data from AI assistants via MCP.

★ 509 · Tools 26 Compare with this →

Firecrawl MCP Server 75 · B

Firecrawl's official MCP server for web search, scraping, and structured extraction for AI agents

★ 7.6k · Tools 11 Compare with this →

BioMCP 75 · B

One binary. One grammar. Evidence from the biomedical sources you already trust.

★ 654 · Tools 11 Compare with this →

Compartment 75 · B

Encrypted, fully offline long-term memory for AI agents

★ 579 · Tools 14 Compare with this →

Compare with other MCP servers

Source revision 3922af856d78 Data synced 2026-10-12 Read the FMRS scoring method