Analytics MCP server for Claude Code, Cursor and Codex
Connect an analytics MCP server to Claude Code or Cursor. Your agent installs cookieless tracking, checks the first event and reads your stats.
Sep 23, 2026 · 4 min
What is an analytics MCP server?#
The Model Context Protocol (MCP) is an open standard that lets AI assistants call tools. An analytics MCP server exposes your analytics as tools: your coding agent can ask how many people visited, which pages they read, where they came from, and which events they triggered, without you opening a dashboard.
Most analytics MCP servers stop there: they read data from a product that is already installed. The hard part for a new project is the install itself. Shipabase’s MCP server covers both: your agent creates the app, adds the SDK in the right file, runs the app and checks that the first event arrived.
Add the Shipabase MCP server#
Pick your agent. The server is remote (Streamable HTTP), so there is nothing to install or run on your machine.
claude mcp add --transport http shipabase https://api.shipabase.dev/mcpThen type /mcp in Claude Code and sign in.
{ "mcpServers": { "shipabase": { "url": "https://api.shipabase.dev/mcp" } }}Then sign in from Cursor’s MCP settings.
gemini mcp add --transport http shipabase https://api.shipabase.dev/mcpThen run /mcp auth shipabase in Gemini CLI.
codex mcp add shipabase --url https://api.shipabase.dev/mcpcodex mcp login shipabaseThe login opens your browser.
The first time, your browser opens Shipabase: sign in, check the agent’s name and click Allow. The agent gets its own token. You can disconnect it anytime in Settings → AI agents.
Your first prompt: add analytics to this app#
Open your project in the agent and ask:
Add Shipabase analytics to this app and track signups.- 1
list_apps, then create_app
The agent looks for an existing app, or creates one named after your project, and gets its public key (
SB-EU-…). - 2
get_setup
It receives the exact steps for your framework: Next.js, React, Vue, Svelte, Node or React Native. Install
@shipabase/js, put the key in an env variable, callinitonce. - 3
Edit the code
It adds
initin the right place and atrackEvent("signup_completed")where signups happen. You review the diff. - 4
verify_install
It runs the app, opens a page and checks that Shipabase received the first event. Page views are tracked automatically, so the first one arrives on load.
Ask questions about your traffic#
Once events come in, the same agent reads your stats. A few prompts that work well:
| Ask | Tool the agent uses |
|---|---|
| How did last week go? | get_overview: visitors, visits, events and time on site, with the week before |
| Which pages do people read most? | get_top with path |
| Do AI assistants send me visitors? | get_top with referrer: ChatGPT, Perplexity, Claude and others are labelled |
| Which browsers and countries? | get_top with engine_name or country |
| Are signups tracked correctly? | get_events: every event with its count, visitors and property keys |
Analytics MCP servers compared#
Several analytics tools now ship an MCP server. What differs is the setup and what the agent is allowed to do. Here is what public comparisons reported in 2026:
| Tool | MCP server | What your agent can do |
|---|---|---|
| Shipabase | Official, remote, OAuth sign-in | Create apps, get the setup, verify the install, read stats |
| Google Analytics 4 | Official, runs locally | Read reports. Needs Python, pipx, two Google Cloud APIs and credentials |
| PostHog | Official, OAuth | Read a very wide product suite: analytics, flags, experiments, errors… |
| Amplitude | Official, OAuth | Read product analytics: charts, cohorts, experiments |
| Plausible | Community servers only | Read stats through an API key |
Pick the MCP server of the tool you already use. If you’re starting a new project and want your agent to do the install too, that’s the gap Shipabase fills.
Privacy, limits and cost#
- No personal data. Shipabase never stores IP addresses or user agents, and the SDK uses no cookies or localStorage. The agent only sees aggregates.
- Limited access. The agent can read stats, list apps and create new ones. It can’t delete an app, rotate a key or change settings.
- Rate limits. 60 calls a minute per agent, 10 new apps a day. Stats answers are cached for a minute, so repeated questions are instant.
- Short answers. Top 10 rows by default and compact JSON, so the agent keeps its context for your code.
- Free. The MCP server is included in every plan, including free.
FAQ
What is an analytics MCP server?
A server that exposes your analytics as tools through the Model Context Protocol, so AI assistants like Claude Code, Cursor, Gemini CLI or Codex can read your stats, and with Shipabase, set up tracking too.
Do I need an API key?
No. The first time, your agent opens the browser and you sign in to Shipabase. The agent then gets its own token, which you can revoke in Settings → AI agents.
Can the agent delete my data?
No. It can read stats, list apps and create new ones. Deleting an app, rotating a key or changing settings stays in the dashboard.
Which frameworks can the agent set up?
Next.js, React (Vite), Vue, Svelte, Node.js and React Native / Expo. Page views are tracked automatically in the browser.
Does it work with other MCP clients?
Yes. Any client that supports remote MCP servers over Streamable HTTP with OAuth can connect to https://api.shipabase.dev/mcp.
Lovable
Bolt