MCP server

The Parsely Model Context Protocol (MCP) server provides AI coding agents with direct, structured access to Parsely’s content across all your sites.

What is the Parse.ly MCP server?

It is a hosted endpoint that exposes your Parse.ly data as a set of tools an AI assistant can call on your behalf. Connect once, authorize with your existing Parse.ly login, and the assistant can then read your analytics — scoped to exactly the sites and permissions your account already has.

• Endpoint: https://dash.parsely.com/mcp
• Transport: Streamable HTTP (JSON-RPC 2.0)
• Authentication: OAuth 2.1 with PKCE — no API keys to copy or paste

Why use it?

• No setup beyond a login. Authorize in your browser with the Parse.ly account you already have. Access is scoped to your existing site permissions.
• Your numbers, in the assistant. The same metrics you see in the dashboard — views, visitors, engaged minutes — answered conversationally.
• Built for AI. Tools and responses are structured so assistants can chain them: find a post, then pull its referrers, then compare to last month.

What can you do with it?

• Site overview — top-line views, visitors, and engaged minutes for today, this week, or this month.
• Real-time — a live pulse of who’s reading right now: concurrent visitors, top pages, and where they’re coming from.
• Traffic trends — daily, weekly, and monthly trends with month-over-month comparison.
• Top content — rank your top posts, videos, and pages by any metric over any time window.
• Content search — find specific posts, videos, and other pages; break down by author, section, or tag.
• Post detail — deep analytics for a single post: its timeline, referrers, and engagement over time.
• Referrers — see where your traffic comes from across channels.
• Conversions — rank conversion labels by metric to see which content drives your goals.
• Campaigns — query and search campaign performance.
• Saved groups — list and manage aspect groups.
• Docs search — ask product questions answered from Parse.ly’s own documentation.

Every tool runs with your permissions — the assistant can only see sites your account can already access.

Connecting

ChatGPT

Since ChatGPT is the most popular AI assistant, it’s the fastest way to try the Parse.ly MCP server. On a plan that supports custom connectors, open Settings → Connectors → Add custom connector, give it a name such as “Parse.ly,” and paste the server URL:

https://dash.parsely.com/mcp

ChatGPT opens your browser to authorize — log in to Parse.ly and click Authorize. Then enable the Parse.ly connector in the chat composer and ask your analytics questions. Custom connectors require a paid ChatGPT plan with connectors enabled by your workspace admin.

Claude Code

claude mcp add --transport http parsely https://dash.parsely.com/mcp

Or add it to ~/.claude.json:

{
  "mcpServers": {
    "parsely": { "type": "http", "url": "https://dash.parsely.com/mcp" }
  }
}

On first use, Claude Code opens your browser to authorize. Log in to Parse.ly and click Authorize.

Claude Desktop

Claude Desktop talks to MCP servers over stdio, so it needs the mcp-remote bridge and Node.js 18 or newer. Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "parsely": {
      "command": "/absolute/path/to/node/bin/npx",
      "args": ["mcp-remote", "https://dash.parsely.com/mcp/"],
      "env": {
        "PATH": "/absolute/path/to/node/bin:/usr/local/bin:/usr/bin:/bin"
      }
    }
  }
}

Use absolute paths — Claude Desktop does not load your shell profile, so a bare npx may pick an old Node and fail. Restart Claude Desktop (Quit, not just close the window) after editing.

Other clients

Any client that supports a remote MCP server over streamable HTTP can connect to https://dash.parsely.com/mcp. Clients that only support stdio can use mcp-remote as shown above.

FAQ

How is my data protected?

You authorize with your normal Parse.ly login over OAuth 2.1. The server never sees a password, and every request is scoped to the sites and permissions on your account. No API secrets are shared.

Does it support Streamable HTTP?

Yes. The endpoint uses MCP’s streamable HTTP transport with JSON-RPC 2.0.

I’m getting authentication errors after about an hour.

Access tokens last one hour. Most clients refresh automatically; if yours does not, reconnect the server. For Claude Desktop, clear the cached token (rm -rf ~/.mcp-auth) and restart.

Claude Desktop fails with “ReadableStream is not defined”.

The mcp-remote bridge is running under Node.js below 18. Point command and env.PATH at a Node 18+ binary, clear the npx cache (rm -rf ~/.npm/_npx/*), and restart.

The authorization screen never appears.

Make sure you are logged in to Parse.ly in the same browser first, then retry the connection.