The Pendo Model Context Protocol (MCP) server lets you connect external AI tools to your Pendo data. With it, you can query analytics, retrieve guides, and surface Pendo insights from AI clients like Claude, Cursor, and ChatGPT.
Note: To share your feedback and help shape what we build next, fill out our Pendo MCP Survey.
Use cases
The Pendo MCP server makes product and customer data available in AI clients where you already work. This lets you ask questions, build agents, or add live Pendo product data in LLM-based clients.
Connect to the Pendo MCP server to:
- Access visitor and account metadata from Pendo.
- Query application analytics and user behavior.
- Find and analyze specific Pages, Features, and Track Events.
- Run event-level aggregation queries on recent usage.
- Analyze visitor activity and engagement patterns.
- Search for and retrieve guides and ideas.
- List AI agents, use cases, and issues.
- Retrieve development logs for a visitor or session.
These capabilities support tasks like preparing for customer calls, investigating adoption or churn signals, enriching support tickets, and powering agents or assistants with live product context. For example, you might ask about your most visited Pages, top features by usage, or behavior patterns within a specific segment.
You can also connect an AI support agent that surfaces Pendo walkthrough links in response to "how do I…?" questions. For more information, see Deliver guides through your AI agent (beta).
Tip: We recommend using dashboards and reports in the Pendo UI for historical trends, advanced reporting, or in-depth guide performance metrics.
MCP versus Leo
The Pendo MCP server and Leo are separate ways to work with Pendo data using AI:
- Leo is Pendo's built-in AI experience. You use Leo inside Pendo to ask questions and get curated insights.
- The Pendo MCP server lets you connect external AI tools, like Claude, Cursor, or ChatGPT, directly to your Pendo data and use it alongside other sources (for example, Google Drive or Atlassian products). You query Pendo from the AI tool you already use, and today, the MCP server is a better fit for complex or high-volume workflows you run from external AI tools.
Use the MCP server when you want to bring Pendo data into multi-source AI workflows or support high-volume, automated analysis from external tools you're already working in.
Use Leo when you want an interactive AI experience for guided exploration and quick questions inside Pendo. For more information, see Analyze product usage with Leo.
How it works
The Pendo MCP server is a remotely hosted service that connects your Pendo subscription data to AI clients that support the Model Context Protocol (MCP). It acts as a remote gateway that makes Pendo analytics data available to AI clients through MCP tool calls.
When connected, the server:
- Authenticates users using OAuth through your Pendo login.
- Accepts requests over Streamable HTTP, with an automatic fallback to Server-Sent Events (SSE).
- Makes Pendo data queryable through standardized MCP tool calls.
Once connected, the MCP server provides real-time access to visitor and account metadata, Page and Feature usage, event summary data, and segment definitions.
The Pendo MCP server can also be used by AI agents and bots that connect programmatically rather than interactively. For example, Pendo and Intercom customers can connect Intercom Fin to Pendo MCP so it can surface guides or use session context during support conversations. For more information, see Serve Pendo guides using Intercom Fin and Improve Fin resolution rate with session context.
Read-only and write tools
The Pendo MCP server provides two categories of tools:
- Read-only tools (enabled by default). Query and retrieve Pendo data. For example, list visitors, retrieve Page and Feature usage, or find guides.
- Write tools (admin opt-in). Take actions in Pendo. Currently includes creating feedback items.
A subscription admin enables each category separately in Settings > Subscription settings > AI access. The read-only tools toggle must be enabled before write tools can be turned on.
To see the full list of tools available, see Pendo MCP tools for AI agents.
Authentication and access
The Pendo MCP server supports two authentication methods, each with different access behavior:
- Interactive OAuth. A Pendo user signs in through an AI client like Claude, Cursor, or ChatGPT. A single connection covers all Pendo subscriptions you belong to within the same region. To query a specific subscription, include its subscription ID in your prompt. You don't need to reconnect or switch subscriptions to query different ones.
- Service accounts. An AI agent or backend integration authenticates programmatically without a person signing in. Each service account is scoped to a single Pendo subscription and the applications within it. To work with multiple subscriptions, create a separate service account in each one. For more information, see Authenticate to the Pendo MCP server with a service account.
Pendo data is regionally isolated, so neither authentication method can access data across regions. If you have Pendo subscriptions in multiple regions, you must configure a separate connection for each region.
Next steps
When you're ready to connect, see Connect to the Pendo MCP server for full setup instructions, including supported AI clients and regional URLs.
Related articles
- Connect to the Pendo MCP server. Set up the connection between Pendo and your AI client.
- Authenticate to the Pendo MCP server with a service account. Configure non-interactive access for agents and backend jobs.
- Pendo MCP tools for AI agents. Learn which tools are available and what data they return.
- Deliver guides through your AI agent. Use MCP to surface Pendo guides in response to support questions.
- Serve Pendo guides using Intercom Fin. Surface Pendo guides in Fin conversations through MCP.
- Improve Fin resolution rate with session context. Give Fin access to Pendo session data during support conversations.