Mcp

MCP Server

Connect AI assistants to your blog via Model Context Protocol

MCP Server

LxBlog exposes a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, Windsurf, and other MCP-compatible clients to interact with your blog content directly. The MCP server runs on the same deployment — no separate service needed.

What is MCP?

The Model Context Protocol is an open standard that lets AI applications discover and use tools exposed by external services. Instead of copying article content into prompts manually, an AI assistant can call LxBlog tools to list, search, and read your articles in real time.

Endpoint

POST/api/mcp

MCP JSON-RPC endpoint. Accepts standard MCP protocol messages with Bearer token authentication.

Headers

Name	Type	Description
`Authorization`*	string	Bearer token: "Bearer lxb_at_..."
`Content-Type`*	string	Must be "application/json"

Authentication

The MCP endpoint uses the same OAuth Bearer token as the REST API. Include your access token in the Authorization header of every request. The token determines which blog the AI assistant can access.

Access tokens expire after 1 hour. If a tool call returns an authentication error, refresh your token using the /api/oauth/token endpoint with your refresh token, then update your MCP client configuration.

Configuration

Claude Desktop

Add the following to your Claude Desktop configuration file (claude_desktop_config.json):

claude_desktop_config.json

{
  "mcpServers": {
    "lxblog": {
      "url": "https://lxblog.app/api/mcp",
      "headers": {
        "Authorization": "Bearer lxb_at_your_access_token"
      }
    }
  }
}

Claude Code

Add the MCP server via the CLI:

Terminal

claude mcp add lxblog \
  --transport http \
  --url https://lxblog.app/api/mcp \
  --header "Authorization: Bearer lxb_at_your_access_token"

Cursor

In your project's .cursor/mcp.json:

.cursor/mcp.json

{
  "mcpServers": {
    "lxblog": {
      "url": "https://lxblog.app/api/mcp",
      "headers": {
        "Authorization": "Bearer lxb_at_your_access_token"
      }
    }
  }
}

Local Development

For local development, replace the URL with your local lxblog instance:

{
  "url": "http://localhost:3000/api/mcp",
  "headers": {
    "Authorization": "Bearer lxb_at_your_access_token"
  }
}

Available Tools

list_articles

List published blog articles with pagination. Returns article metadata, content, tags, and cover images.

Parameters

Name	Type	Description
`page`	integer	Page number (starts at 1)(default: `1`)
`limit`	integer	Articles per page (max 100)(default: `20`)

Example response

{
  "articles": [
    {
      "id": "clx...",
      "slug": "getting-started-with-ai",
      "title": "Getting Started with AI",
      "description": "A beginner's guide...",
      "content": "# Getting Started with AI\n\n...",
      "coverImageUrl": "https://...",
      "tags": ["ai", "tutorial"],
      "publishedAt": "2025-01-15T10:00:00.000Z",
      "contentHash": "a1b2c3..."
    }
  ],
  "pagination": { "page": 1, "limit": 20, "total": 12, "totalPages": 1 }
}

get_article

Get a single published article by its slug or CUID. Returns the full article including markdown content.

Parameters

Name	Type	Description
`identifier`*	string	Article slug (e.g. "my-article") or CUID

search_articles

Search published articles by keyword or tag. Matches against title, description, and tags.

Parameters

Name	Type	Description
`query`	string	Search query (matches title and description)
`tag`	string	Filter by exact tag name
`limit`	integer	Maximum results to return (max 50)(default: `10`)

sync_articles

Diff-based article sync. Send your known articles (id + contentHash) and receive only new, updated, and deleted articles. Send an empty array for initial sync to get all articles.

Parameters

Name	Type	Description
`knownArticles`	array	Articles you already have. Each object needs "id" and "contentHash" fields.(default: `[]`)

Example response

{
  "new": [{ "id": "...", "slug": "new-article", "title": "...", ... }],
  "updated": [{ "id": "...", "slug": "changed-article", "title": "...", ... }],
  "deleted": ["old-article-id"],
  "syncedAt": "2025-01-15T12:00:00.000Z"
}

get_blog_info

Get blog metadata including name, description, language, article count, and connection status. No parameters required.

Example response

{
  "blog": {
    "id": "clx...",
    "name": "My Tech Blog",
    "slug": "my-tech-blog",
    "description": "A blog about technology",
    "language": "en",
    "articleCount": 42
  },
  "connection": {
    "connectedAt": "2025-01-01T00:00:00.000Z",
    "lastSyncAt": "2025-01-15T12:00:00.000Z"
  }
}

Usage Examples

Asking Claude about your blog

Once configured, you can ask your AI assistant natural language questions and it will use the LxBlog tools automatically:

Example prompts

"List all my published articles"
→ Uses list_articles

"What articles do I have about machine learning?"
→ Uses search_articles with query "machine learning"

"Show me the full content of my article about React hooks"
→ Uses search_articles then get_article

"How many articles does my blog have?"
→ Uses get_blog_info

"What articles have been updated since my last sync?"
→ Uses sync_articles with your cached article hashes

Testing with cURL

You can test the MCP endpoint directly with cURL by sending JSON-RPC messages:

Terminal

# Initialize the MCP session
curl -X POST https://lxblog.app/api/mcp \
  -H "Authorization: Bearer lxb_at_your_token" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "initialize",
    "params": {
      "protocolVersion": "2024-11-05",
      "clientInfo": { "name": "test", "version": "1.0.0" },
      "capabilities": {}
    }
  }'

# List available tools
curl -X POST https://lxblog.app/api/mcp \
  -H "Authorization: Bearer lxb_at_your_token" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 2,
    "method": "tools/list"
  }'

# Call a tool
curl -X POST https://lxblog.app/api/mcp \
  -H "Authorization: Bearer lxb_at_your_token" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 3,
    "method": "tools/call",
    "params": {
      "name": "list_articles",
      "arguments": { "page": 1, "limit": 5 }
    }
  }'

Scopes

MCP tools respect the same OAuth scopes as the REST API:

articles:read — Required for list_articles, get_article, search_articles, and sync_articles
blog:read — Sufficient for get_blog_info

The MCP server is stateless — each request is independent. There are no sessions to manage or connections to keep alive. This makes it ideal for serverless deployments on Vercel.

Blog Metadata

SDKs & Libraries