MCP Server AGENTS.md

A unified MCP server for managing agent definition Markdown files (rules, workflows, subagents, claude.md, cursorrule, windsurfrule, etc.). Write once, use everywhere.

Note

AGENTS.md published by OpenAI.

Warning

This is just a prototype project.

Overview

mcp-server-agents-md is a lightweight MCP server that centralizes all your agent definition files in Markdown format. It allows you to easily share and reuse rules, workflows, subagents, and other agent specs across different code agent platforms like Claude, Cursor, Windsurf, and more.

• Serve Markdown-based agent definitions via MCP.
• Easy integration with any MCP-compliant agent environment.

Features

Tip

To help LLMs more quickly identify tool invocation commands, we've designed a trigger command (cc:, carbon copy to agent), similar to / and @ in other products. For example, to trigger an agent workflow that generates a git commit message, you can write cc:commit . Multiple trigger commands can exist, defined in the frontMatter.trigger field of each agent's markdown file.

Agents (workflows/rules):

• commit,git-commit - Create well-formatted commits with conventional commit messages and emojis.
• fast-commit,fast-git-commit - Generate 3 commit message suggestions based on the staged changes.
• issue-report,analyze-issue - Fetch GitHub issue details and create a comprehensive implementation specification.
• bugfix,code-bugfix - Streamline bug fixing workflow from issue creation to pull request.
• clean,code-clean - Fix all code formatting and quality issues in the entire codebase.
• check,code-check - Perform comprehensive code quality and security checks.
• code-analysis,code-report - Perform advanced code analysis with multiple inspection options.
• create-docs - Create comprehensive documentation for specified components or features.
• update-docs - Generate LLM-optimized documentation with concrete file references and flexible formatting.
• changelog,add-to-changelog,update-changelog - Update the project's CHANGELOG.md file with a new entry.
• task,implement-task - Approach task implementation methodically with careful planning and execution.
• mermaid,create-mermaid - Generate Mermaid diagrams for visualizing code structure and relationships.

Agents, workflows, rules source:

• Agent Rules

Usage

To use with AI client apps such as Claude, VSCode, Cursor, Windsurf, Cline, Cherry Studio, Chatbox, etc., add the MCP server configuration below.

On Mac system:

{
  "mcpServers": {
    "Agents": {
      "command": "npx",
      "args": [
        "--registry=https://registry.npmjs.org/",
        "-y",
        "mcp-server-agents-md@latest"
      ]
    }
  }
}

On Windows system:

{
  "mcpServers": {
    "Agents": {
      "command": "cmd",
      "args": [
        "/c",
        "npx",
        "--registry=https://registry.npmjs.org/",
        "-y",
        "mcp-server-agents-md@latest"
      ]
    }
  }
}

Next, you need to select an AI model that supports tool calling (Tools Call/Function Call) in your client application and enter simple prompts, such as:

• git commit - Natural language request
• Call the tool to generate commit information - Explicit tool request
• cc:commit - Trigger command, recommended
• cc:fast-commit - Trigger command, recommended

Development

Install dependencies:

npm install

Build the server:

npm run build

Start the MCP server:

npm run start

Publish

Version releases use the standard-version tool.

npm run release
# or Release as a Pre-Release
npm run release -- --prerelease alpha

License

MIT