@docsyncdev/mcp-server

MCP (Model Context Protocol) server for DocSync - enables AI coding assistants like Claude Code and Cursor to search and interact with your documentation.

Features

• Context-Aware Docs: Get documentation for any source file with full hierarchy (unit → topic → module)
• Semantic Search: Search your documentation using natural language queries
• Documentation Browser: List and browse docs by level (module, topic, unit)
• Full Document Retrieval: Get complete document content by ID
• Compact Mode: Condensed output for quick lookups (reduces token usage ~30%)
• Caching: Built-in LRU cache reduces API calls and improves latency
• Rate Limiting: Server-side rate limiting prevents abuse

Quick Start

1. Get Your API Key

1. Go to your DocSync dashboard at docsync.dev
2. Navigate to Settings > API Keys
3. Click Create API Key and copy the key

2. Get Your Project ID

1. In DocSync dashboard, go to Settings > API Keys
2. Select your project from the dropdown
3. Copy the project ID from the generated config

3. Configure Your AI Assistant

Claude Code

Add to your claude_desktop_config.json or .claude/settings.json:

{
  "mcpServers": {
    "docsync": {
      "command": "npx",
      "args": ["-y", "@docsyncdev/mcp-server"],
      "env": {
        "DOCSYNC_API_KEY": "ds_your_api_key_here",
        "DOCSYNC_PROJECT_ID": "your-project-uuid"
      }
    }
  }
}

Cursor

Add to your Cursor MCP configuration:

{
  "mcpServers": {
    "docsync": {
      "command": "npx",
      "args": ["-y", "@docsyncdev/mcp-server"],
      "env": {
        "DOCSYNC_API_KEY": "ds_your_api_key_here",
        "DOCSYNC_PROJECT_ID": "your-project-uuid"
      }
    }
  }
}

Available Tools

search_docs

Search documentation semantically. Returns relevant chunks matching your query.

Parameters:

• query (string, required): Natural language search query
• limit (number, optional): Maximum results (default 10, max 50)
• compact (boolean, optional): Return condensed output with just paths and scores (default: false)

Example:

How does authentication work in this project?

Compact mode example:

search_docs({ query: "authentication", compact: true })

→ **5 results for "authentication"**
  - docs/topics/auth.md (95%)
  - docs/modules/security.md (82%)
  - ...

get_docs_for_file ⭐

The killer feature. Get documentation relevant to a source code file, with full context.

When you're working on src/auth/jwt.ts, this returns:

• Unit doc (JWT implementation details) - full content
• Module doc (Security architecture) - full content
• Related topics based on concepts (Auth, Tokens, etc.) - summaries only

One call for immediate context. Related topics give you paths to dive deeper if needed.

Parameters:

• file_path (string, required): Path to the source code file
• include_content (boolean, optional): Include full content (default: true)
• compact (boolean, optional): Return condensed output for quick lookups (default: false)

Example:

Get docs for src/services/payment-processor.ts

Returns:

# 📄 Unit Documentation
## PaymentProcessor Class
Detailed API for the payment processor...
[full content]

# 🏗️ Module Context
## Billing & Payments
Overview of the billing system architecture...
[full content]

# 📚 Related Topics (3)
- **Payment Processing** (payments)
  Path: docs/topics/payments.md
  How payments flow through the system

- **Validation** (validation)
  Path: docs/topics/validation.md
  Input validation patterns

- **Error Handling** (errors)
  Path: docs/topics/errors.md
  Error handling strategies

Compact mode returns summaries with topic list:

get_docs_for_file({ file_path: "src/auth/jwt.ts", compact: true })

→ ## src/auth/jwt.ts
  **JWT Handler**
  Handles JWT token creation and validation

  - **Module:** Security — Security patterns

  **Related Topics (2):**
  - Authentication (authentication): docs/topics/auth.md
  - Token Management (tokens): docs/topics/tokens.md

list_docs

Browse all documentation, optionally filtered by level.

Parameters:

• level (string, optional): Filter by "module", "topic", "unit", or "all" (default)
• compact (boolean, optional): Return condensed output with just titles and paths (default: false)

Example:

List all module-level docs

Compact mode returns a simple list without summaries:

list_docs({ level: "module", compact: true })

→ **3 docs in MyProject**

  **Module** (3)
  - Authentication: docs/modules/auth.md
  - Billing: docs/modules/billing.md
  - API: docs/modules/api.md

Environment Variables

| Variable | Required | Description | | -------------------- | -------- | ----------------------- | | DOCSYNC_API_KEY | Yes | Your DocSync API key | | DOCSYNC_PROJECT_ID | Yes | Your DocSync project ID |

Rate Limits

• search_docs: 200 requests per hour
• get_docs_for_file: 200 requests per hour
• list_docs: 200 requests per hour

Rate limits are per API key. If you hit a rate limit, the tool will return an error message.

Caching

The MCP server includes an in-memory LRU cache:

• Cache size: 100 entries per cache type
• TTL: 5 minutes
• Repeated queries return cached results instantly

Cache clears when the MCP server process restarts.

Troubleshooting

"DOCSYNC_API_KEY environment variable is required"

Make sure you've added the env section to your MCP configuration with your API key.

"Rate limit exceeded"

Wait a few minutes before making more requests. Consider if your agent is making too many searches.

"Document not found"

The document ID may be stale if documentation was recently updated. Run a new search to get current document IDs.

Connection issues

1. Make sure your config uses npx with the -y flag: "args": ["-y", "@docsyncdev/mcp-server"]. The -y flag lets npx auto-install the package without prompting, which is required for MCP stdio communication.
2. Check that your API key is valid in DocSync settings
3. Verify your project ID is correct
4. Ensure you have internet connectivity