Fullstory Model Context Protocol (MCP)

AI agent · Coding agent · Engineering & development by Fullstory

Connect your AI to live behavioral context for automated session review and investigation.

Fullstory MCP is a specialized interface that connects AI agents and LLMs like Claude and Gemini to Fullstory's behavioral data. It enables engineering and product teams to automate session reviews and technical investigations directly from their terminal or IDE. Choose it to scale behavioral analysis and resolve user friction using AI-native tools and structured event data.

Agent specs

The full spec card (context, pricing per token, modalities and more) on AntSpace. Sign up free to see the full spec card.

Launched
2026-11-10
Languages
English

Category: AI Agents

Channels
API, Terminal, IDE, In-app
Editors and tools
Claude Code, Cursor, Codex, Devin, Terminal
Models
Claude, Claude Code, Codex, Cursor, Gemini Enterprise
Deployment
Cloud
Protocol Standard
Model Context Protocol (MCP)
Supported LLMs
Claude, Claude Code, Gemini Enterprise, Codex
  • Generally available as of November 10, 2026
  • Native connectors for Claude, Cursor, and Gemini Enterprise
  • Enables Agentic Session Review via terminal or IDE
  • Close to 10 million tool calls made during beta period
  • Reduces investigation time from hours to minutes

Overview

Fullstory Model Context Protocol (MCP) is a specialized interface that connects third-party AI agents and Large Language Models (LLMs) to Fullstory's behavioral data. Designed for engineering, product, and data teams, it enables "Agentic Session Review," allowing AI coding agents like Claude Code, Cursor, or Gemini to reason over structured behavioral truth directly from their own environment. This eliminates the need for humans to manually watch linear replays by providing agents with a semantic event map of user interactions. The protocol includes a purpose-built toolset for progressive data disclosure, such as review-open and review-zoom, which agents use to investigate technical failures and user friction. By integrating with tools like Sentry and Datadog, Fullstory MCP allows agents to move directly from a flagged error to the specific session that explains it. It is generally available as of November 2026 and has already supported millions of tool calls during its beta phase.

Agentic Session Review (ASR)

  • Terminal-Based Session Access: AI coding agents like Claude Code or Cursor can open real production sessions directly from the terminal to investigate issues.
  • Semantic Event Mapping: Agents read a semantic event map of user interactions to reason over structured behavioral truth without linear playback.
  • Agent-Native Toolset: The protocol provides purpose-built tools like review-open, review-zoom, and review-snapshot for progressive disclosure of session data.
  • State Comparison: Agents can compare application state across different timestamps to explain exactly what went wrong in a digital experience.

AI Ecosystem Connectivity

  • Claude & Claude Code Integration: Native connectors for Anthropic's Claude and Claude Code provide seamless behavioral context to LLMs.
  • Gemini Enterprise Connector: Direct connection to Google Gemini Enterprise equips the model with live user behavior data for analysis.
  • Developer Tool Support: The protocol supports AI-native IDEs and coding assistants including Cursor, Codex, and Devin.
  • Ecosystem Linking: Session URLs can be attached to signals from Sentry, Datadog, and Segment to give agents a direct path to the relevant session.

Intelligence & Automation

  • Natural Language Interaction: Users can discover frustrations, measure behavior, and have AI read sessions using natural language queries.
  • Automated Investigation: The protocol allows teams to automate technical investigations and session reviews at a scale impossible for humans.
  • Rapid Connection: Teams can connect their AI tools to Fullstory data in just a few clicks using native connectors.

What it does

  • Review customer sessions at scale
  • Investigate experience problems
  • Trace issues to root causes
  • Explain technical failures
  • Compare application state across timestamps
  • Identify user frustrations
  • Measure digital behavior

Published results

  • Close to 10 million tool calls made during beta
  • Weekly MCP tool calls grew 10x in Q2
  • Reduces investigation time from hours to minutes
  • Enables evidence-ranked product proposals in minutes instead of weeks

Use cases

  • Engineering teams use AI coding agents to open production sessions from the terminal and diagnose root causes of errors.
  • Product designers use AI agents to review hundreds of sessions per week and generate evidence-ranked product proposals.
  • Support teams attach session URLs to Sentry or Datadog signals so AI agents can automatically investigate flagged issues.
  • Data teams equip LLMs with live behavioral context to drive automated actions and personalize user experiences.

Specifications

Channels
API, Terminal, IDE, In-app
Integrations
Claude, Claude Code, Codex, Cursor, Gemini Enterprise, Sentry, Datadog, Segment, LangSmith, Intercom, Linear, LaunchDarkly, Snowflake, Fin
Editors and tools
Claude Code, Cursor, Codex, Devin, Terminal
Human control
AI agents investigate and explain issues while humans spend their time on final decisions instead of searching and analyzing.
Models
Claude, Claude Code, Codex, Cursor, Gemini Enterprise
Deployment
Cloud
Protocol Standard
Model Context Protocol (MCP)
Supported LLMs
Claude, Claude Code, Gemini Enterprise, Codex
Supported IDEs
Cursor, Claude Code, Codex, Devin
Primary Interface
Terminal / API-based Agentic Review
Data Capture Type
Semantic Event Mapping
Integration Ecosystem
Sentry, Datadog, Segment, LangSmith, Intercom, Linear, LaunchDarkly
Availability Status
Generally Available (GA) as of Nov 10, 2026

How it works

  1. Connect AI Tools: Teams connect their preferred AI tools like Claude or Gemini to Fullstory using native MCP connectors in a few clicks.
  2. Access Session Data: AI agents access real production sessions directly from the terminal or IDE using the Model Context Protocol.
  3. Automated Investigation: The agent reads semantic event maps and uses tools like review-zoom to investigate specific moments of friction.
  4. Review Insights: The AI agent delivers an evidence-ranked explanation or proposal, allowing humans to make final decisions.

FAQ

What is Fullstory MCP?

Fullstory MCP is a specialized interface that connects third-party AI agents and LLMs, such as Claude or Gemini, to Fullstory's live behavioral data.

What is Agentic Session Review?

Agentic Session Review is a new category of behavioral capture built specifically for AI agents as the primary user, allowing them to read semantic event maps rather than watching linear replays.

Which AI tools and models does Fullstory MCP support?

It supports native connectors for Claude, Claude Code, Codex, Cursor, and Gemini Enterprise, as well as integrations with tools like Sentry and Datadog.

How can I access Fullstory MCP?

Fullstory MCP is generally available as of November 10, 2026, and is typically used in conjunction with StoryAI Premium or enterprise-level plans.

Added by AntSpace from public sources (fullstory.com).

Related

More AI agents by Fullstory