Skip to main content

Fullstory MCP

The Fullstory MCP server connects your AI assistant directly to your Fullstory product data. Using the Model Context Protocol, it lets any MCP compatible AI client (Cursor, Claude Desktop, Claude Code, and others) query your Fullstory data, analyze user behavior, and surface insights in natural language, without leaving your development environment.

Connect once, then ask questions like:

  • "What are the biggest frustrations affecting our checkout page right now?"
  • "How many users hit a network error on the payment page last month?"
  • "Show me sessions where users abandoned the signup form."
Looking for Fullstory for Customer Agents?

This guide covers the Fullstory MCP server for analytics—answering natural-language questions about your Fullstory behavioral data. If you instead want to give AI support and sales agents context about individual sessions, see Fullstory for Customer Agents.

Prerequisites​

Before connecting an MCP client, confirm the following for your Fullstory org:

  • MCP Beta access is enabled for your account. Fullstory MCP is in private beta—sign up for the waitlist if you haven't been onboarded yet.
  • StoryAI features are enabled. This is required. If StoryAI is opted out for your org, MCP will appear to connect but return zero tools. See the note below.
StoryAI features must be enabled

Fullstory MCP requires StoryAI features to be enabled for your org. If your org has opted out of StoryAI, your MCP client will appear to connect successfully (OAuth completes, the server shows as "connected") but no tools will be available.

An account Admin can verify or change this at Settings > StoryAI Features in Fullstory. If you don't see that page, you don't have the role needed to change it. Ask an Admin in your org.

See FAQ & Troubleshooting if your client is connected but no tools appear.

Getting started​

The fastest way to get started depends on your AI client.

Any other MCP client—Add the server manually:

  1. Set up your AI client with the server URL.
  2. Authenticate with Fullstory using one of:
    • OAuth (default)—On first connection, your client opens a browser to sign in. See Authentication.
    • API key—Best when a per-user browser sign-in isn't practical (for example, a gateway or shared service). See Authentication.
  3. Start asking questions about your product data.

See the client setup guides for step-by-step instructions for specific clients.

Go further​

Once you're connected, the tools reference and example workflows below show what's possible.

Tools​

See the Tools Reference for a full list of available tools and what each one returns.

Example Workflows​

The following end-to-end examples show common patterns for getting the most out of the Fullstory MCP server.

  • Find what's frustrating users
    Automatically surface rage clicks, dead clicks, errors, crashes, and form abandons. Discover the biggest problems across your product or zoom into a specific page, funnel, or user segment.

  • Measure user behavior
    Ask natural language questions to get real numbers. "How many users hit this error last week?" returns a count, trend, or breakdown without writing a query.

  • Let AI watch the sessions for you
    Surface real sessions for any signal or behavior pattern and have AI read the full event transcript. Get a plain-language summary of what users did, where they struggled, and why, without watching a single recording yourself.

  • Friction signal to session investigation
    Start from a friction signal, find affected sessions, and have AI read the event transcripts—all in one multi-turn conversation. Walk out with a root-cause summary and a session link, no recording required.

  • Build and refine a metric
    Define a custom metric in plain language, iterate on the definition, and get a trend with a breakdown—without leaving your AI client.

  • Release impact
    Compare user behavior before and after a release. Confirm improvements landed, catch regressions, and quantify the delta—in a single multi-turn workflow.