Integrating AI development agents into your workflow requires structured interfaces that allow models to query, read, and write documentation context efficiently. docmd satisfies this need via a native Model Context Protocol (MCP) server and an extensible Agent Skills database.

Model Context Protocol (MCP) Setup

The Model Context Protocol connects LLM environments directly to your local workspace tools.

1. Claude Desktop Integration

Add the following to your desktop configuration file (typically at ~/Library/Application Support/Claude/claude_desktop_config.json on macOS or %APPDATA%\Claude\claude_desktop_config.json on Windows):

{
  "mcpServers": {
    "docmd": {
      "command": "npx",
      "args": ["@docmd/core", "mcp"],
      "cwd": "/path/to/your/docs/project"
    }
  }
}

2. IDE Integration (Cursor / Windsurf)

In your editor’s MCP settings panel, add a new server using the stdio transport:

• Command: npx @docmd/core mcp
• Transport: stdio

Available MCP Tools

Once connected, the following tools become available to the agent:

1. search_docs(query): Performs a workspace-wide full-text search.
2. read_doc(route): Retrieves the raw Markdown contents of a specific route.
3. validate_docs(): Lints the entire documentation and returns validation errors (e.g., broken links).
4. get_llms_context(): Fetches the consolidated llms-full.txt context file.

Leveraging Agent Skills (SKILL.md)

When you run docmd init in your project, the engine automatically generates a SKILL.md file in your root workspace. This file serves as a prompt instruction card for any AI agent working on your repository.

Best Practices for AI Agents

1. Read SKILL.md First: Instruct your agents to read the SKILL.md file at the start of a coding session. This teaches the model about custom Callouts, OpenAPI markup, and file structures.
2. Validate After Edits: Whenever an agent modifies Markdown files, it should call the validate_docs tool (or run npx @docmd/core validate) to verify that no relative links or anchor paths are broken.
3. Synchronize Locales: If the project uses versioning or multiple languages, agents should use the comparison matrix to ensure all translations stay in sync.