da-mcp

A remote Model Context Protocol (MCP) server for Document Authoring (DA). This server provides LLM assistants like Claude or ChatGPT with direct access to DA management operations.

Features

• 10 DA Admin Tools: Complete set of tools for managing DA repositories
• Remote Access: Deployable on Cloudflare Workers with global edge distribution
• Streamable HTTP: Modern MCP transport protocol for remote servers
• Token Pass-through: Simple authentication by passing DA API tokens through Authorization header
• Production Ready: Error handling, logging, CORS support, and health checks
• TypeScript: Fully typed codebase for reliability and maintainability

Architecture

┌─────────────────┐
│   MCP Client    │
│ (Claude/Cursor) │
└────────┬────────┘
         │ Streamable HTTP
         │ + DA Token
         ↓
┌─────────────────────────┐
│  DA MCP                 │
│  ┌──────────────────┐   │
│  │   MCP Server     │   │
│  │   (10 Tools)     │   │
│  └──────────────────┘   │
└───────────┬─────────────┘
            │ HTTPS + Token
            ↓
┌─────────────────────────┐
│  DA Admin API           │
│  (admin.da.live)        │
└─────────────────────────┘

Project Structure

src/
├── index.ts              # Cloudflare Worker entry point
├── mcp/
│   ├── server.ts         # MCP server initialization
│   ├── tools.ts          # Tool definitions (schemas)
│   └── handlers.ts       # Tool implementation handlers
└── da-admin/
    ├── client.ts         # DA Admin API client
    └── types.ts          # TypeScript types

Available Tools

Tool	Description
da_list_sources	List sources and directories in a repository
da_get_source	Get content of a specific source file
da_create_source	Create a new source file
da_update_source	Update an existing source file
da_delete_source	Delete a source file
da_copy_content	Copy content between locations
da_move_content	Move content between locations
da_get_versions	Get version history for a file
da_lookup_media	Lookup media references
da_lookup_fragment	Lookup fragment references

Prerequisites

• Node.js 18+ and npm
• Cloudflare account (free tier works)
• Wrangler CLI installed (npm install -g wrangler)
• DA Admin API token

Installation

1. Clone and install dependencies:

git clone <repository-url>
cd da-mcp
npm install

2. Configure Wrangler:

Edit wrangler.toml if needed to customize your deployment settings.

Development

Local Development

Run the server locally with hot reload:

npm run dev

The server will be available at http://localhost:8787

Test Endpoints

• Health check: http://localhost:8787/health
• MCP endpoint: http://localhost:8787/mcp

Testing with MCP Inspector

1. Start the local server: npm run dev
2. Open MCP Inspector
3. Configure connection:
   • Type: Streamable HTTP
   • URL: http://localhost:8787/mcp
   • Headers: Authorization: Bearer YOUR_DA_TOKEN

Deployment

Deploy to Cloudflare Workers

# Deploy to production
npm run deploy

# Or deploy to development environment
wrangler deploy --env development

Public URLs

After deployment, your MCP server is accessible at:

• Direct MCP Endpoint:
  https://da-mcp.franklin-prod.workers.dev/mcp

• IMS-Authenticated via AEM API Router:
  https://mcp.adobeaemcloud.com/adobe/mcp/da
  (Supports Adobe IMS login. See the Authentication section below for details.)

Client Configuration

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or equivalent:

{
  "mcpServers": {
    "da-live-admin": {
      "type": "streamable-http",
      "url": "https://da-mcp.adobeaem.workers.dev/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_DA_USER_IMS_TOKEN"
      }
    }
  }
}

VS Code / Cursor

Add to .vscode/mcp.json or Cursor settings:

{
  "mcpServers": {
    "da-admin-mcp-direct": {
      "url": "https://da-mcp.adobeaem.workers.dev/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_DA_USER_IMS_TOKEN"
      }
    },
    "da-prod-mcp": {
      "url": "https://mcp.adobeaemcloud.com/adobe/mcp/da"
    }
  }
}

Authentication

The server uses simple token pass-through authentication:

1. Client sends DA Admin API token in the Authorization header
2. Server extracts the token and passes it to DA Admin API
3. All requests to DA Admin API use this token

Authorization Header Format:

Authorization: Bearer YOUR_DA_USER_IMS_TOKEN

or simply:

Authorization: YOUR_DA_USER_IMS_TOKEN

Note: If you are accessing the API through the public (authenticated) URL of the API router, IMS (Adobe Identity Management Service) login is automatically handled by the AEM API router. In this case, you do not need to provide a DA Admin API token in the Authorization header—the IMS login flow will provide authentication for you.

Usage Examples

Once configured, you can ask your AI assistant to perform DA operations:

Claude, can you list all the sources in the adobe/my-docs repository?

Please get the content of docs/index.md from the adobe/my-docs repository.

Create a new file at docs/new-page.md with some markdown content.

API Endpoints

GET /health

Health check endpoint returning server status.

Response:

{
  "status": "healthy",
  "service": "da-mcp",
  "version": "1.0.0",
  "environment": "production",
  "timestamp": "2025-01-07T12:00:00.000Z"
}

POST /mcp

MCP protocol endpoint for tool execution. Requires Authorization header with DA Admin API token.

Error Handling

All tools include comprehensive error handling:

• 401 Unauthorized: Missing or invalid DA Admin token
• 404 Not Found: Invalid endpoint
• 408 Timeout: Request took longer than 30 seconds
• 500 Internal Error: Server-side errors with details

Errors are formatted for easy understanding by LLM clients.

Logging

The server logs important events and errors to Cloudflare Workers logs:

# View logs in real-time
wrangler tail

# View logs for specific environment
wrangler tail --env production

Monitoring

Monitor your deployed Worker:

1. Cloudflare Dashboard: View invocations, errors, and performance
2. Wrangler Tail: Real-time logs (wrangler tail)
3. Health Endpoint: Regular health checks at /health

Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Make your changes with tests
4. Submit a pull request