Replayful
Sign inGet started

MCP server for AI clients

Connect Claude Code, Claude Desktop, or any MCP client to your Replayful workspace and ask questions about your own data in natural language.

What you need

A workspace API key. Create one on the API keys page. The endpoint is https://replayful.co/api/mcp/mcp (streamable HTTP), authenticated with the key as a Bearer token. The API keys page has copy-paste setup snippets for the clients below.

Connect Claude Code

Run this in your terminal, swapping in your own key:

claude mcp add --transport http replayful https://replayful.co/api/mcp/mcp --header "Authorization: Bearer <YOUR-KEY>"

Connect Claude Desktop

Claude Desktop doesn't speak streamable HTTP directly, so it connects through mcp-remote. Add this to your mcp.json:

{
  "mcpServers": {
    "replayful": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://replayful.co/api/mcp/mcp",
        "--header",
        "Authorization: Bearer <YOUR-KEY>"
      ]
    }
  }
}

What the AI can do

Once connected, the AI can:

  • List your projects
  • Pull analytics overview, acquisition, behaviour, and realtime data for a project
  • List sessions
  • Export a single session or a visitor's stitched journey as an AI-ready timeline
  • List goals
  • Create a goal

Everything is scoped to the key's workspace. All tools are read-only except creating a goal, and exports are the derived timeline, never raw recording streams.

Unlimited on every paid plan

Questions asked over MCP do not count against the monthly AI question allowance that applies to the in-app Ask tab, and they never will. The inference runs on your own AI account rather than ours, so there is nothing for us to meter.

That makes MCP the better choice if you work in an AI client all day, or want an agent querying your analytics on a schedule. The in-app tab exists for the times you are already in Replayful and just want an answer. Two tools available over MCP are deliberately not exposed in the tab: session export, because a single timeline is far larger than a chat answer needs, and creating a goal, because the tab is read-only by design.