pieowl

Feature · MCP

MCP: ask your AI about your users.

Let the AI assistant you already use dig into your Pieowl data. Skip dashboards and reports when you just want a quick answer.

  • 40+ tools
  • Works with Claude, Cursor & Windsurf
  • Nothing to install
Claude · connected to PieowlDemo data

Which funnel step has the biggest drop-off this week?

pieowl · get_funnel · last 7 days

Most people leave at Checkout → Payment: 41% of those who start checkout don’t enter payment details.

  • Viewed a product1,015100%
  • Added to cart81280%
  • Started checkout59058%
  • Entered payment34834%
Updated just now

The basics

What is an analytics MCP server?

4 steps to start
  1. 01Create a token
  2. 02Copy your token
  3. 03Add the URL to your AI client
  4. 04Start asking questions

The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to outside data and tools. An analytics MCP server is the bridge: it turns your assistant's requests into analytics queries and returns data it can reason about.

Without it, getting an answer looks like this: open the dashboard, find the right view, build the chart, read the number, switch back to what you were doing. If the answer raises another question, repeat.

With Pieowl's MCP server, you ask and get the answer inside the tool you're already in.

  • "What's the week-4 retention for users who signed up last month?"
  • "Which funnel step has the biggest drop-off this week?"
  • "Show me the last 5 sessions for user@example.com"

Pieowl hosts the server at https://api.pieowl.com/mcp, so there's nothing to install or run on your machine. In Settings → Developers → AI assistants, click Create token and copy the ready-made config into your AI client. That page has setup steps for Claude Desktop, Claude Code, Cursor, Windsurf, VS Code and Raycast.

The server has more than 40 tools that cover:

  • Events and sessions: query events, filter by property, look at single sessions
  • User profiles and groups: find users, see their history and metrics
  • Funnels and retention: run a funnel or a cohort retention table for any events
  • Pages and traffic: top pages, entry and exit pages, referrers, countries and devices
  • Reports and dashboards: read saved reports, and (with a root key) create or update reports and dashboards for you
  • Google Search Console: clicks, impressions, top queries and SEO opportunities (when you've connected Search Console)

The MCP server can't send events or change user profiles.

There is no chat inside Pieowl: you ask in your own AI assistant. In the dashboard, the AI command bar sets the date range and filters from what you type.

MCP on every plan

What you can do with it

More than 40 tools across your analytics, all usable in plain language from any MCP client.

MCP · 6 capabilities
  • Ask questions in plain language

    No SQL and no chart builder. Describe what you want to know and your AI runs the query. Works for funnels, retention, top pages, user lookups and more.

  • Funnel and retention analysis

    Run a funnel for any sequence of events or pull a cohort retention table without opening the dashboard. Ask a follow-up and the AI queries again.

  • User and group lookup

    Find a user by email or ID and see their events, sessions and metrics. For B2B products, look up groups and their members.

  • Google Search Console data

    If Search Console is connected to your site, your AI can also ask about top queries, ranking opportunities and per-page performance.

  • Build reports and dashboards

    With a root key, ask your AI to create or update a report or dashboard, and open it in Pieowl afterwards. It can't send events or edit user profiles.

  • Stay where you work

    Analytics questions come up while you build or debug. MCP keeps you in your editor, so you get the answer without switching tabs.

Set up

How to connect your AI to Pieowl

Nothing to install. The MCP server is hosted at api.pieowl.com/mcp.

  1. 1

    Create a token

    Open Settings → Developers → AI assistants and click Create token. It makes a read key for that site. A root key can see every site in your team.

  2. 2

    Copy your token

    The token is shown once, already in a ready-made config. It's your key ID and secret joined by a colon and base64-encoded, so you can also build it yourself.

  3. 3

    Add the URL to your AI client

    Add https://api.pieowl.com/mcp as an MCP server and send the token as an Authorization: Bearer header. Clients that can't set headers can pass it as ?token= in the URL instead.

  4. 4

    Start asking questions

    Ask about users, funnels, retention or events. Not sure which events you have? Ask the AI to list them first.

Use cases

Who uses it

Anyone who has analytics questions while doing something else.

  • Founders and product teams

    Someone in a meeting asks about last week's drop-off. You ask Claude, get the answer in seconds, and move on.

  • Developers in Cursor or Windsurf

    You're fixing a checkout bug and need to know how many users hit that step. Ask in your editor and get the answer next to your code.

  • Teams without an analyst

    Most small teams don't have someone who lives in the dashboard. MCP lets anyone on the team get answers without learning the tool.

FAQ

Frequently asked questions

Common questions about Pieowl's MCP server.

Which AI clients work with Pieowl MCP?

Any client that can connect to a remote MCP server over HTTP. The settings page has setup steps for Claude Desktop, Claude Code, Cursor, Windsurf, VS Code and Raycast.

Do I need to install anything?

No. Pieowl hosts the MCP server at https://api.pieowl.com/mcp. You add that URL and your token to your AI client. Nothing runs on your machine.

How does authentication work?

Your token is your client ID and secret joined by a colon, base64-encoded. Send it as an Authorization: Bearer header. If your client can't set headers, add it to the URL as ?token= instead. That works the same, but a token in a URL is easier to leak into logs and history.

What's the difference between a read client and a root client?

A read client can only see the site it was created in. A root client can see every site in your team. Use read for one site; use root if you want to ask across sites.

Can the AI change my data?

It can create, update and delete reports and dashboards when you ask it to. It has no tools to send events, edit user profiles or change site settings.

Your first dashboard in five minutes.

14-day free trial, no card needed. Cloud in the EU. No cookies.