@kinlab/kin-mcp

Superseded by @kinlab/kin, the canonical Kin install surface. It includes the MCP server (kin mcp start) along with the full CLI. This package keeps working for existing configurations, but new setups should install @kinlab/kin.

@kinlab/kin-mcp is the npm-friendly launcher for Kin's MCP server (the installed command is still kin-mcp).

It downloads the matching Kin release archive from GitHub, verifies the published SHA-256 checksum, extracts both the kin CLI and the kin-daemon it depends on into a local cache, and runs:

kin mcp start

On first run it:

• provisions kin-daemon next to kin and points the CLI at it with KIN_DAEMON_BIN, so the MCP path never depends on a locally-built kin, a stale daemon on PATH, or a pre-existing daemon;
• defaults KIN_MCP_TOOL_PROFILE=agent-default, so agents see the small curated tool surface instead of the full internal one;
• starts (or reuses) the repo daemon automatically, and only then serves tools;
• requires an explicit kin init (no silent init) unless you opt in with KIN_MCP_AUTO_INIT=1.

If a runnable Kin cannot be provisioned (unsupported target, release download blocked), the wrapper exits with a precise guided fix instead of a stack trace.

Usage

Use it directly from an MCP client with npx:

{
  "mcpServers": {
    "kin": {
      "command": "npx",
      "args": ["-y", "@kinlab/kin-mcp"]
    }
  }
}

To pin a version, add it to the package spec. Take the number from the published version list rather than from this page, because a version written into a README is stale by the next release and pins whoever copies it to whatever shipped when the line was typed:

{
  "mcpServers": {
    "kin": {
      "command": "npx",
      "args": ["-y", "@kinlab/kin-mcp@<version>"]
    }
  }
}

Requirements

• Node.js 20+
• macOS, Linux, or native Windows x64
• A Kin-initialized repository (kin init)

The native Windows archive carries semantic vector search but does not include transparent filesystem projection. Use WSL2 when you need projection.

Cache

The wrapper caches the downloaded kin and kin-daemon binaries under:

• macOS: ~/Library/Caches/kin-mcp
• Linux: ~/.cache/kin-mcp
• Windows: %LOCALAPPDATA%\kin-mcp\Cache

Set KIN_MCP_CACHE_DIR to override the cache location.

Environment

• KIN_MCP_KIN_BINARY: run a specific local kin binary instead of downloading one (you are then responsible for its kin-daemon)
• KIN_BINARY_PATH: alias for KIN_MCP_KIN_BINARY
• KIN_MCP_CACHE_DIR: override the cache directory
• KIN_MCP_AUTO_INIT=1: allow the wrapper to run kin init . when .kin/ is missing
• KIN_MCP_TOOL_PROFILE: override the default agent-default tool profile
• KIN_MCP_RELEASE_BASE_URL: override the GitHub release download base URL

Local Check

npx -y @kinlab/kin-mcp --print-bin         # provisioned kin path
npx -y @kinlab/kin-mcp --print-daemon-bin  # provisioned kin-daemon path

Then initialize a repository and let the MCP client launch kin-mcp.

First-run smoke proof

test/smoke-first-run.mjs exercises the whole first-run path against built Kin binaries: it stages kin + kin-daemon into a throwaway cache (no pre-existing daemon, no dev PATH state), runs the wrapper, and drives the MCP stdio protocol through one safe semantic tool (kin_graph_status). Run it with:

cargo build --release -p kin-cli -p kin-daemon
KIN_BIN=target/release/kin \
KIN_DAEMON_BIN=target/release/kin-daemon \
npm run smoke