kilter-mcp

An unofficial MCP server that lets your AI agent read and analyse your own Kilter Board logbook.

Add it to Kiro (or another MCP client), give it your Kilter login, and then just talk to the agent:

How did my last Kilter session go? What are my current projects at 30°? Compare my climbing at 20° and 40°. Show my grade pyramid. What are my hardest sends? Am I progressing?

Read-only. Nothing is written to your Kilter account.

Add to Kiro + Other Clients (Claude, Codex, Cursor etc.)

Click the button (Kiro shows a confirmation dialog first), then jump to step 3. Or do it by hand:

1. Open your MCP config in Kiro: command palette → Kiro: Open user MCP config (JSON) (or the workspace one, .kiro/settings/mcp.json).
2. Add the server:

{
  "mcpServers": {
    "kilter": {
      "command": "uvx",
      "args": ["kilter-mcp"],
      "env": {
        "KILTER_USERNAME": "${KILTER_USERNAME}",
        "KILTER_PASSWORD": "${KILTER_PASSWORD}"
      },
      "disabled": false,
      "autoApprove": []
    }
  }
}

3. Provide your Kilter credentials. Either:
   • set KILTER_USERNAME and KILTER_PASSWORD as environment variables (Kiro expands ${VAR} after you approve the variables in Settings → Mcp Approved Env Vars; see how to set them on macOS, Linux and Windows), or
   • replace the ${...} placeholders with the literal values. Only do this in the user config (~/.kiro/settings/mcp.json), never in a workspace config you might commit.
4. Save. Kiro starts the server and the kilter_* tools appear. Ask away.

Requirements: uv installed (uvx comes with it). Python is handled by uv.

Kiro CLI users can do the same with kiro-cli mcp add --name kilter --scope global --command uvx --args kilter-mcp --env KILTER_USERNAME=... --env KILTER_PASSWORD=....

Other clients

kilter-mcp is a standard local MCP server, so it also works in Claude Desktop, Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI and Zed. Copy-paste configs for each are in docs/clients.md. (ChatGPT's chat interface only accepts remote HTTP servers, so it is not supported directly; details in the same document.)

Example questions

Show my latest Kilter session.
What did I climb last Saturday?
List my sends from this month at 40°.
What are my current projects?
Which projects have I put the most attempts into?
Show my grade pyramid at 30°.
What is my flash rate per grade?
What are my five hardest sends?
Compare my climbing at 20° and 30°.
How has my climbing progressed over the last six months?

Tools the agent gets

All tools are read-only and only ever see the account whose credentials you configured.

How the data is interpreted

• One logbook entry is one climb at one wall angle on one date, with the number of tries in that entry. Re-logging a climb creates a new entry.
• status is flash (topped first try), send (topped) or attempt (not topped).
• A project is a (climb, angle) pair with attempts but no top at that angle.
• Grades are Kilter's current consensus grade for the climb at that angle. grade is the V-scale and font_grade the Font scale; the raw difficulty_id (1-39) is always included. The grade table is fetched from Kilter at startup with an embedded fallback.

Security and privacy

• Your username and password are only used to obtain an OAuth token from Kilter's login server (idp.kiltergrips.com). They never leave your machine otherwise.
• Tokens are kept in memory for the life of the MCP process and are never written to disk.
• Credentials and tokens are never included in tool output, logs or error messages.
• The server makes requests only to idp.kiltergrips.com and portal.kiltergrips.com.
• Prefer the ${KILTER_PASSWORD} form so the password lives in your shell environment or secret manager rather than in a JSON file.

Disclaimer

kilter-mcp is an unofficial community project and is not affiliated with or endorsed by Kilter Grips. It talks to the same endpoints the Kilter app uses; these are not a documented public API and may change or stop working at any time. Kilter's Terms of Use restrict access to their services outside the official apps; use this project at your own discretion, only with your own account, and keep request volume low (the server fetches your logbook once per tool call).

Contributing

Issues and pull requests are welcome. See CONTRIBUTING.md for the development setup (git clone, uv sync, uv run pytest), architecture notes, and how to add a tool.

License

MIT