Real product teams are using Pendo MCP to get instant behavioral insights, validate product decisions in minutes, and build better products — all from a plain-language prompt. Here's everything we covered and how to start today.
Six months ago, MCP (Model Context Protocol) was a brand-new concept. Today, it feels like every product team is talking about it — and the smartest ones are already using it to move faster, make better decisions, and build products their users actually love.
Last week, we brought together hundreds of product leaders for a hands-on MCP workshop. We heard from a hackathon winner who built a real customer health dashboard in a weekend. We did live demos. We answered the questions you've been too afraid to ask. Here's what you need to know.
Pendo MCP lets you ask plain-language questions of your product data inside Claude, ChatGPT, Gemini, or Cursor — without ever logging into Pendo. It's read-only, OAuth-secured, and available to every Pendo customer today.
Think of MCP as the USB-C for AI. Just a few years ago, every device had a different cord — lightning, mini-USB, proprietary connectors. USB-C standardized all of that. MCP is doing the same thing for AI tools and data sources.
Model Context Protocol is an open standard developed by Anthropic (the company behind Claude). It acts as a universal, secure bridge between your AI tools and external data sources. For Pendo customers, that means your LLM can now securely access your behavioral data, NPS responses, guide completions, and feature adoption metrics — in real time — from whatever AI tool you already use.
Stephanie Tanzler, VP of Product at Auto Titling Company, had never used MCP before Pendo's hackathon. She walked away with a working customer health dashboard that her account management team is now actually using. Here's how she did it.
"I can build features using Pendo data with a single natural language prompt. Claude Code understood Pendo's data structure immediately — it was like, this is the future." — Stephanie Tanzler, VP of Product, Auto Titling Company
Her problem was simple but painful: her account management team couldn't tell which customers were at risk without digging through Pendo themselves. The churn signals were buried. The insights weren't actionable.
Her solution: use Claude Code + Pendo MCP to pull login frequency, feature adoption events, and vehicle registration completions — then build a red/yellow/green health score based on what retained customers actually looked like after 12+ months.
MCP gave Stephanie a way to understand what data to use before building a full API integration. She could explore, validate, and prototype with real data — then decide whether it was worth investing in production infrastructure. That's a fundamentally new way to de-risk product decisions.
Not everyone is ready to build a full dashboard on day one. Here's the framework our team walked through live:
We have 50+ pre-built prompts you can copy-paste directly into Claude or your preferred LLM today. Browse the full prompt library — filter by use case (activation, retention, feature adoption, competitive intel) and start getting answers in minutes. No setup required beyond connecting MCP.
Claude Projects are dedicated workspaces where you give the AI standing instructions — which Pendo sub to pull from, how to format output, what to prioritize. Set it up once, share it with your team, and everyone benefits from consistent, high-quality analysis. Add your Google Drive or Confluence connector and it gets even more powerful.
Connect Pendo behavioral data with win/loss docs from Google Drive and in-flight PRDs from Jira/Confluence — all in one conversation. The result: a Q1 planning brief that cites your own data, competitive intelligence, and engineering context simultaneously. This is where MCP gets genuinely unfair.
The walkthrough our team did live illustrated exactly what "run" looks like in practice. The setup: a PM heading into a Q1 planning session who needs to make the case for where to invest in AI features. Here's the prompt that kicked it off:
From that single prompt, the agent surfaced that onboarding guides had zero completions, dashboard customization engagement was critically low, and a key setup action was essentially absent from user flows.
Then — without leaving the conversation — it pulled win/loss data from Google Drive showing that 6 of 13 recent losses cited onboarding as a contributing factor. It cross-referenced Jira and Confluence to confirm no existing PRD was addressing new user activation. And it synthesized everything into a one-page executive brief with specific feature recommendations, data citations, and competitive context.
This isn't just querying a dashboard. It's combining quant (Pendo events), qual (win/loss docs), and operational context (PRDs) in a single reasoning loop. That used to take a week of cross-functional coordination. Now it takes a conversation.
Pages, features, track events, NPS, guide polls, Listen feedback — all queryable via MCP today.
Build and publish Pendo guides directly from your LLM conversation, no Pendo UI required.
Connect MCP to your own product's AI assistant or support bot. Serve Pendo guides inside external chat interfaces — show, don't just tell.
Query the visualizations and reports you've already built in Pendo — directly through MCP.
