Skip to content
MCP server

Kitsune MCP

Hot-reload an MCP server you are building and mount community servers mid-session

Runs locallyStatus and history
Install
Pricing
Open source
Platforms
CLI
Runs
Locally
Version
0.23.0
Listing updated
3 days ago

About Kitsune MCP

Kitsune MCP is an open-source MCP server for developer tools. It runs locally on your machine and is started by your AI client.

The problem Kitsune solves is narrow: an agent edits an MCP tool, but the client keeps running the old process until someone restarts it. Here connect, call and reload are themselves tools, so the agent can start a server, change the code, call reload on it and re-run the tool in the same session.

The second function is borrowing. A search tool queries 7 registries covering 130,000+ servers, shapeshift mounts a single tool rather than a whole server, and shapeshift with no arguments drops it again. An auth tool can add an API key mid-session. When Docker is installed, npm and PyPI community packages run in a locked-down container; your own server run through connect is a plain local process and does not need Docker.

The site is candid about when to skip it: with only 1 to 3 trusted servers, configure them directly. Without client tool deferral it costs roughly 1,774 tokens per turn, and a cold mount takes 1 to 15 seconds. It needs Python 3.12 or newer and uv, and installs with uvx kitsune-mcp.

Install

Run in a terminal

claude mcp add kitsune-mcp --env SMITHERY_API_KEY='<SMITHERY_API_KEY>' -- npx -y kitsune-mcp

Add to claude_desktop_config.json

{
  "mcpServers": {
    "kitsune-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "kitsune-mcp"
      ],
      "env": {
        "SMITHERY_API_KEY": "<SMITHERY_API_KEY>"
      }
    }
  }
}

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "kitsune-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "kitsune-mcp"
      ],
      "env": {
        "SMITHERY_API_KEY": "<SMITHERY_API_KEY>"
      }
    }
  }
}

Add to .vscode/mcp.json

{
  "servers": {
    "kitsune-mcp": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "kitsune-mcp"
      ],
      "env": {
        "SMITHERY_API_KEY": "<SMITHERY_API_KEY>"
      }
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "kitsune-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "kitsune-mcp"
      ],
      "env": {
        "SMITHERY_API_KEY": "<SMITHERY_API_KEY>"
      }
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "kitsune-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "kitsune-mcp"
      ],
      "env": {
        "SMITHERY_API_KEY": "<SMITHERY_API_KEY>"
      }
    }
  }
}

Configuration

  • SMITHERY_API_KEYOptional. Adds Smithery to registry search; the other registries need no key. Get one at smithery.ai/account/api-keys
  • SMITHERY_API_KEYOptional. Adds Smithery to registry search; the other registries need no key. Get one at smithery.ai/account/api-keys

Tools this server provides (6)

Taken from the project documentation; the running server may differ.

connect
Starts your own MCP server as a local process.
shapeshift
Mounts a server's tools, or just selected ones, into the session; with no arguments it drops them.
call
Calls a mounted tool.
reload
Restarts a connected server on fresh code and remounts it.
search
Searches MCP registries for servers.
auth
Saves a credential such as an API key mid-session.

What it does

Hot reload
Restarts your server on fresh code and remounts it without restarting the client.
Registry search
Searches 7 registries covering 130,000+ servers.
Single-tool mounting
Mounts one tool from a server instead of all of them.
Docker cage
Runs npm and PyPI community packages in a hardened container when Docker is present.
Mid-session credentials
Adds an API key during a session.
Version pins and confirmation
Requires explicit confirmation to run community code.

Good for

  • Let an agent fix and retest its own MCP tool without a client restart
  • Try a scraping server once without editing config
  • Run an unfamiliar PDF tool package in a container
  • Develop a local server with connect and reload

Pricing

Open source

Free; optional Smithery key for extra registry search

Prices change. Check kaiser-data.github.io before you buy.

Kitsune MCP FAQ

Do I need Docker?

No. Hot-reload of your own server never needs it; Docker only cages community packages.

What are the requirements?

Python 3.12+, uv, and node/npx for npm-based servers.

Which clients does it work with?

Claude Code, Claude Desktop, Cursor, Windsurf, Cline, Zed and others.

Does it cost tokens?

Without tool deferral, about 1,774 tokens per turn, per the site.

Is a key required?

No. SMITHERY_API_KEY is optional and only adds Smithery to registry search.

Is Kitsune MCP free?

Yes. Kitsune MCP is open source, so you can use and self-host it for free.

How do I add Kitsune MCP to Claude Code?

Run this command in your terminal: claude mcp add kitsune-mcp --env SMITHERY_API_KEY='<SMITHERY_API_KEY>' -- npx -y kitsune-mcp

What are the best alternatives to Kitsune MCP?

The closest alternatives to Kitsune MCP are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

Alternatives to Kitsune MCP

  1. REA

    MCP server

    CLI and MCP server that lets coding agents reverse engineer apps and binaries

    Runs locally
    • Open source
    • 464K GitHub stars
    • Developer tools
  2. Codebase Memory

    MCP server

    Local MCP server that builds a code knowledge graph for AI coding agents

    Runs locally
    • Open source
    • 48.7K GitHub stars
    • Knowledge and memory
  3. GitHub

    MCP server

    Official MCP server connecting AI assistants to GitHub repos, issues and workflows

    Not checked yet
    • Remote
    • Open source
    • 34.5K GitHub stars
    • Developer tools
  4. MCP toolkit giving coding agents symbol-level code retrieval and editing

    Runs locally
    • Open source
    • 31.5K GitHub stars
    • Developer tools
  5. XERJ

    MCP server

    Local search engine for AI agents with cited answers over your code and docs

    Runs locally
    • Open source
    • 19.2K GitHub stars
    • Knowledge and memory
  6. Ouroboros

    MCP server

    Spec-first layer that pins requirements and verifies results of AI coding agent work

    Runs locally
    • Open source
    • 6.4K GitHub stars
    • Developer tools