Cartograph

Name: Cartograph
Author: benteigland11

@benteigland11

リポジトリ

Cartograph について

Local First MCP for making your AI generated code reusable and improvable across projects.

#vibecoding #reuse #store #memory

基本情報

カテゴリ

その他

トランスポート

stdio

公開者

benteigland11

投稿者

Ben Teigland

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "cartograph": {
      "command": "cartograph-mcp",
      "args": [],
      "description": "Local-first reusable code registry for AI agents",
      "homepage": "https://github.com/benteigland11/cartograph-mcp"
    }
  }
}

ツール

ツールは検出されませんでした

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is Cartograph?

Cartograph is an MCP server that exposes a focused set of Cartograph widget workflow tools for AI agents. It provides search, inspect, install, create, validate, check-in, and configuration actions without mirroring the entire CLI, keeping the agent-facing surface compact and task-oriented.

How to use Cartograph?

Install via pip install cartograph-mcp, then configure it as an MCP server in your client (e.g., Claude Desktop, Claude Code, Codex, Gemini CLI, or Cursor) by pointing to the cartograph-mcp command. The server delegates to the underlying cartograph-cli for all operations.

Key features of Cartograph

Exposes registry widget actions: search, inspect, install, rate
Manages installed widgets: upgrade, uninstall
Provides widget status, creation, validation, and check-in tools
Reads and writes core Cartograph configuration defaults
Groups tool surface around agent intents rather than CLI mirroring
Falls back to the full CLI for administrative and recovery operations

Use cases of Cartograph

Search and inspect reusable widgets before writing new logic
Install, upgrade, or uninstall widgets within a project
Create, validate, and check in new widgets with a dry-run pipeline
Adjust daily workflow defaults like auto-publish, visibility, and governance

FAQ from Cartograph

How does Cartograph differ from the full Cartograph CLI?

The MCP intentionally exposes only the daily widget workflow surface (search, install, create, validate, check-in, config). Administrative, recovery, and cloud operations remain in the cartograph CLI.

What are the runtime dependencies of Cartograph?

The cartograph-mcp package depends on cartograph-cli and shells out to it as the source of truth for all command execution.

How is data isolation handled during testing or development?

Integration tests isolate HOME, XDG_CONFIG_HOME, XDG_DATA_HOME, XDG_CACHE_HOME, WIDGET_LIBRARY_PATH, and the project working directory so they do not touch real widget libraries or user configuration.

What configuration defaults does Cartograph expose?

The cartograph_config tool exposes auto-publish, visibility, governance, cloud, show-unavailable, and publish-registry settings, read and written through the CLI's --json path.

How is Cartograph tested?

It has two test layers: command-contract tests that mock the CLI runner and verify exact commands, and isolated integration tests that run the real Cartograph CLI in a temporary environment.