.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
Ask Claude to show you every user on your Pro plan who signed up last month but hasn't triggered your main activation event, and you get an answer in seconds. No SQL, no dashboard filters, no CSV exports. That's the UserGuiding MCP Server: a direct line between your AI tool and the live product data you already track in UserGuiding.
It's built for product, growth, sales, and support teams who want to query their own users the way they talk about them. Before a renewal call, pull a customer's recent activity in one prompt. When a ticket comes in, pull the user's full profile before you reply. When you want to know which cohorts are activating and which are dropping off, ask the question out loud. You skip the dashboard, skip the exports, and get the specific user, attribute, or count you needed.
This post walks through what you can do with it, how to connect Claude, Cursor, or OpenAI Codex in under two minutes, the full tool catalog, and how UserGuiding compares to other product adoption platforms on AI integration.
Key Takeaways
- The UserGuiding MCP Server connects AI tools directly to your UserGuiding project's live data, so you can run natural-language queries against your users, companies, events, and Knowledge Base without SQL or exports.
- Works with Claude (Desktop and Claude.ai), Cursor, OpenAI Codex, and any other MCP-compatible AI tool.
- 17 tools are available across five capability areas: user management, search and analytics, event tracking, company lookup, and Knowledge Base management.
- Available on all UserGuiding plans. No developer help needed to set it up.
- Pairs with UserGuiding's No-Code Event Tracking to cover the analytics stack most teams pay Amplitude, Heap, or Mixpanel for. Track events without code, query them in plain English through the MCP server.
- Get your API key at Panel > Settings > MCP & API, or read the full setup guide in the Knowledge Base.
What is MCP?
MCP (Model Context Protocol) is an open standard released by Anthropic in 2024. It lets AI assistants call external services in a structured, authenticated way, similar to how a developer calls an API, but mediated by the AI model. Instead of pasting data into a chat window or hooking up ad-hoc integrations, the AI tool connects once to an MCP server and gets a well-defined set of tools it can invoke on your behalf.
Claude (Desktop and Claude.ai), Cursor, and OpenAI Codex support MCP natively. That means any MCP server you connect, including UserGuiding's, becomes accessible across your AI workflows: coding sessions, research, product questions, pre-demo prep, and anything else you use those tools for.
What you can do with UserGuiding MCP
The UserGuiding MCP Server turns the product data you already track into something you can query conversationally. Five use cases teams are already running.
User research without SQL. Filter your users by any attribute or event, combine criteria, slice by cohort. No dashboard filter stack, no CSV export. The AI picks the right search tools automatically.
"Show me all users on the Pro plan who signed up in the last 30 days and triggered the export_csv event at least once."Pre-call research for sales. Before a renewal call or a prospect demo, pull everything the user has done in your product: plan, signup date, feature usage, which goals they hit, which events they triggered this week. Walk into the call with real context instead of asking "what plan are you on?"
"Look up user@acme.com. What's their plan, when did they sign up, and what have they done in the product this week?"
Support context. When a ticket comes in, pull the full user profile and recent interaction history in one prompt. Cuts the 2-3 back-and-forth messages that usually start with "what plan are you on?" and "when did you sign up?"
"Pull the full profile for user@example.com so I can see their plan, features used, and recent activity."
Activation and onboarding analysis. Compare cohorts, find differences in behavior, identify the onboarding steps that separate activated users from ones who churn. No dashboard, no saved segments.
"Show me users who completed our onboarding checklist vs. users who dropped off at step 3. What did the completers do that the others didn't?"
Expansion and churn signals. Find users whose usage patterns suggest they're ready to upgrade, or users on paid plans who are going quiet. These are leads sales and CS usually miss because nobody flagged them.
"Find users on the Starter plan who triggered the api_rate_limit_hit event more than ten times this month."How to set it up
Getting started takes two minutes. First, grab your API key from the UserGuiding Panel: Settings > Project Settings > MCP & API. Then drop the server into your AI tool's MCP config.
Server URL:
https://mcp.userguiding.com/mcp/sse
Authentication: pass your API key as a header (UG-API-KEY: <your-api-key>) or as a query parameter for tools that don't support custom headers (?api_key=<your-api-key>).
Claude Desktop
Edit claude_desktop_config.json:
{ "mcpServers": { "userguiding": { "url": "https://mcp.userguiding.com/mcp/sse", "headers": { "UG-API-KEY": "<your-api-key>" } } } }
Cursor
Edit .cursor/mcp.json:
{ "mcpServers": { "userguiding": { "url": "https://mcp.userguiding.com/mcp/sse", "headers": { "UG-API-KEY": "<your-api-key>" } } } }
OpenAI Codex
Codex follows the same MCP server config format as Cursor. Drop the UserGuiding server entry into your Codex mcp.json using the JSON above. Check OpenAI's current Codex MCP docs for the exact file location on your system.
For the full multi-tool setup, including the Claude.ai admin connector flow and query-parameter authentication, see the UserGuiding MCP Server KB article.
What's inside: the full tool catalog
The UserGuiding MCP Server exposes 17 tools across five capability areas. Each is callable in plain English by any MCP-compatible AI assistant.
User Management
get_user: Look up a user by ID. Returns attributes, company, and interaction history. Connects to User Identification, Segmentation, Analytics.upsert_user: Create or update a user. Merges attributes and optionally attaches to a company. Connects to User Identification, Segmentation.delete_user: Permanently delete a user. Connects to User Identification.reset_user_history: Reset a user's interaction history while keeping their custom attributes. Connects to Guides, Checklists, Surveys, Hotspots, Resource Centers.list_users: List all users with cursor-based pagination. Connects to User Identification, Segmentation.
Search and Analytics
search_users: Search and filter users by attributes or events. Connects to Segmentation, Analytics.get_user_count: Count users matching filters without fetching full records. Connects to Segmentation, Analytics.list_attributes: Discover all available user attributes and their data types. Connects to User Identification, Segmentation.list_events: Discover all tracked event names in your project. Connects to Analytics, No-Code Analytics, Custom Alerts.
Events
track_event: Track a named event for a user, optionally with metadata. Connects to Analytics, No-Code Analytics, Custom Alerts, Segmentation.
Companies
get_company: Look up a company by ID. Returns attributes and member user IDs. Connects to Segmentation, account-level targeting.list_companies: List and filter companies with pagination and sorting. Connects to Segmentation, account-level targeting.
Knowledge Base
search_kb_articles: Search KB articles by text query. Returns IDs, titles, descriptions, categories. Connects to Knowledge Base.get_kb_article: Retrieve full KB article content by ID, including title, body, status, and locales. Connects to Knowledge Base.create_kb_article: Create a new KB article with a title, content, and optional description. Connects to Knowledge Base.update_kb_article: Update the title or content of an existing KB article by ID. Connects to Knowledge Base.delete_kb_article: Permanently delete a KB article by ID. Connects to Knowledge Base.
You can combine these in a single prompt. The AI assistant picks the right tools to answer your question.
See it in action
The example in the previous section came out of a real Tuesday. Our CTO connected UserGuiding's MCP server to Claude, ran the ten-prospect query on our own users, and posted the output on LinkedIn. The reactions and replies are worth skimming if you want to see how other product, sales, and growth teams are already thinking about AI-native access to product data. The post below has the full context.
The No-Code Analytics combo
The MCP server gets more interesting when you pair it with UserGuiding's No-Code Event Tracking. Point-and-click on elements in your product to start capturing events. No code, no engineer time, no new SDK to install. Then query those events through the MCP server in the same sentence you'd use to describe the question out loud.
For most product, growth, and CS teams, that pair replaces a stack that used to look like Amplitude, Heap, or Mixpanel for event tracking plus a separate analytics tool for exploration. With UserGuiding, event capture lives in the same platform as your onboarding, your checklists, and your Knowledge Base, and the MCP server turns product analytics into a chat interface inside Claude, Cursor, or Codex.
The result: product analytics without a dedicated analyst, a separate analytics contract, or an engineering backlog.
What the MCP Server gives you
A quick summary of what's inside the box:
- Read and write access. Query users, companies, and events; upsert profiles, track events, and create or update Knowledge Base articles from your AI tool.
- Interaction depth. Individual guide, checklist, survey, and hotspot events with timestamps, not just aggregate counts.
- No-code setup. Connect via claude.ai connector settings or a JSON config file. Takes about two minutes.
- Natural-language queries. Ask about users in plain English. The AI picks the right tools automatically.
- Real-time data. Queries reflect current state, not a cached export.
This doesn't make MCP the only way to work with product data. Dashboards are still faster for glancing at charts, and SQL is still the right tool for deep analytics. What changes is the middle ground: the ad-hoc questions that used to mean pulling a filter or opening a notebook now happen inline in whatever AI tool you're already in.
Get started
Getting started takes one API key and a few lines of config. Find the key at Panel > Settings > MCP & API, drop the server into your AI tool's config, and ask your first question. For the full setup reference, including the Claude.ai admin connector flow and examples of filter syntax, see the UserGuiding MCP Server KB article.
Frequently Asked Questions
Does it work with ChatGPT or other AI tools?
MCP support is currently native in Claude (Desktop and Claude.ai), Cursor, and OpenAI Codex. ChatGPT and Gemini do not support MCP natively as of April 2026. Any tool that adopts MCP as a transport will be able to connect to the UserGuiding server without additional setup on our side.
Is the data real-time?
Yes. MCP queries hit UserGuiding's live database, not a cached export. If a user triggered an event 30 seconds ago, your AI tool sees it.
What about data privacy?
Data stays in your UserGuiding project. The AI tool sends queries to the UserGuiding MCP server and receives results. No bulk export or storage happens on the AI side.
Can users break things with write operations?
Write tools (upsert_user, track_event, delete_user, reset_user_history, and the KB create/update/delete tools) use the same API key as read tools. Treat the key like any admin credential: share it carefully. Read-only use (search, get, list) is safe regardless.
Do we need engineering help to set it up?
No. It's a connection you enable in your AI tool's settings. Total setup time is about two minutes.
Which plans include MCP?
MCP is available on all UserGuiding plans.
How does this compare to Amplitude, Heap, or Mixpanel?
Amplitude, Heap, and Mixpanel are product analytics platforms with their own event-tracking SDKs and query interfaces. UserGuiding's No-Code Event Tracking captures the same kinds of user events without code or engineering help, and the MCP server lets you query them in natural language instead of building dashboards. For product, growth, and CS teams who want analytics without a dedicated analyst, it replaces a separate analytics contract with something already in your UserGuiding plan.
.png)
