About Mitr Analytics
Mitr Analytics is an MCP server for asking questions about a Mitr analytics workspace through an AI client. It provides read access to reports and tracking data, alongside permission-gated tools for managing apps, funnels, QR campaigns, and webhooks.
Key features
- 01App traffic and performance reports
- 02Recent raw event inspection
- 03Funnel conversion and drop-off results
- 04Usage, alerts, and workspace overview
- 05Permission-gated app and campaign changes
Use cases
- 01Compare visitor trends across apps
- 02Find drop-off points in a signup funnel
- 03Check event properties and UTM tracking
- 04Review quota usage and webhook health
MCP support
Connects MCP clients to Mitr through a hosted remote endpoint or a local npx setup. It exposes analytics and workspace tools; write operations depend on permissions granted to the API key.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://mcp.mitranalytics.dev/mcpclaude mcp add --transport http shiv9thakur-mitr https://mcp.mitranalytics.dev/mcp{"mcpServers":{"shiv9thakur-mitr":{"type":"http","url":"https://mcp.mitranalytics.dev/mcp"}}}Open in Cursor{"servers":{"shiv9thakur-mitr":{"type":"http","url":"https://mcp.mitranalytics.dev/mcp"}}}Open in VS Code{"mcpServers":{"shiv9thakur-mitr":{"type":"http","url":"https://mcp.mitranalytics.dev/mcp"}}}{"mcpServers":{"shiv9thakur-mitr":{"serverUrl":"https://mcp.mitranalytics.dev/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
https://mcp.mitranalytics.dev/mcpAuth requiredTools
Details
From the project README@mitr-analytics/mcp
Ask Claude, Gemini or any other MCP client about your Mitr analytics, in plain English. It reads your own data (apps, reports, funnels, alerts, plan usage) and hands it to the assistant to analyse.
Mitr is cookie-free analytics, so there is no consent banner to wire up and nothing here changes that: this server only reads what is already in your dashboard.
What you can ask
- "Which of my apps had the biggest drop in visitors this week, and where did it come from?"
- "Show my signup funnel for the last 30 days. Where do people leave?"
- "Is anything odd in the last events for my marketing site? Are the UTM tags coming through?"
- "How much of my monthly event quota have I used?"
- "Register a new web app called Docs for docs.example.com."
Set up (about two minutes)
You need a Mitr account. Nothing to install if you use the hosted endpoint.
1. Create an API key. In the dashboard open Settings → API keys & AI assistants → Create key, give it a name (for example "Claude Code") and copy it. It is shown once.
Keys are read-only unless you tick permissions when creating them: "Can register apps", "Can manage funnels", "Can manage QR campaigns", "Can manage webhooks", and "Can delete" (which also needs one of the manage permissions). They expire after 365 days by default (30, 90, 365 days or never). A key stops working if you leave or are removed from the workspace it was created in.
2. Connect your assistant.
Hosted (no install)
Point any client that supports remote MCP servers at https://mcp.mitranalytics.dev/mcp and send the key in the Authorization header. Never put the key in the URL; the endpoint refuses it.
Claude Code:
claude mcp add --transport http mitr https://mcp.mitranalytics.dev/mcp --header "Authorization: Bearer mitr_live_..."
Cursor (~/.cursor/mcp.json):
{
"mcpServers": {
"mitr": {
"url": "https://mcp.mitranalytics.dev/mcp",
"headers": { "Authorization": "Bearer mitr_live_..." }
}
}
}
VS Code (.vscode/mcp.json):
{
"servers": {
"mitr": {
"type": "http",
"url": "https://mcp.mitranalytics.dev/mcp",
"headers": { "Authorization": "Bearer mitr_live_..." }
}
}
}
Gemini CLI (~/.gemini/settings.json):
{
"mcpServers": {
"mitr": {
"httpUrl": "https://mcp.mitranalytics.dev/mcp",
"headers": { "Authorization": "Bearer mitr_live_..." }
}
}
}
Local (npx)
For clients that run a command, such as Claude Desktop. You need Node.js 18 or newer. Open claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/, Windows: %APPDATA%\Claude\) and add:
{
"mcpServers": {
"mitr": {
"command": "npx",
"args": ["-y", "@mitr-analytics/mcp"],
"env": { "MITR_API_KEY": "mitr_live_..." }
}
}
}
The same works in Claude Code (claude mcp add mitr --env MITR_API_KEY=mitr_live_... -- npx -y @mitr-analytics/mcp) and the Gemini CLI if you prefer running it locally.
3. Ask a question (restart the assistant first if it's a local setup). The dashboard shows when the key was last used.
Tools
| Tool | What it does |
|---|---|
list_apps | Every app in your workspace, with the siteId the other tools need. |
get_app_report | A 7-day report for one app: visitors, pageviews, traffic, sources, countries, devices, web vitals, conversions. |
get_recent_events | The latest raw events for one app, to check what is really being tracked. |
list_funnels | Your saved conversion funnels. |
get_funnel_results | Step-by-step conversion and drop-off for one funnel over a date range. |
get_alert_rule | Your traffic spike/drop alert settings, if you have set one. |
get_billing_usage | Your plan, subscription status and event usage. |
get_workspace_overview | One summary of every app's traffic, webhooks, QR campaigns and funnels, with anything that needs attention. |
list_qr_campaigns | Your QR campaigns with ids, status, scan totals and your plan's QR limits. |
get_qr_campaign | One campaign with its codes and per-code scan counts. |
get_webhook_health | Per webhook endpoint: delivery counts by status and the latest failure. Payloads are never returned. |
add_app | Register a new app. Needs "Can register apps". The app's secret key is not returned: copy it from the dashboard. |
create_funnel, update_funnel | Create a funnel from ordered page-path or event steps, or replace a funnel's name and steps. Needs "Can manage funnels". |
delete_funnel | Permanently delete a funnel. Needs "Can manage funnels" and "Can delete". |
create_qr_campaign, update_qr_campaign | Create a campaign that sends scanners to an https URL, or rename it, change the destination, or set its status to active, paused or archived. Needs "Can manage QR campaigns". |
add_qr_code | Add a printable code to a campaign, labelled by placement. Needs "Can manage QR campaigns". |
create_webhook, update_webhook | Create an https webhook endpoint, or change its URL and event types, or pause it. The signing secret is not returned: copy it from the dashboard. Needs "Can manage webhooks". |
delete_webhook | Permanently delete a webhook endpoint. Needs "Can manage webhooks" and "Can delete". |
Every change is its own permission on the key, and deletes need "Can delete" as well. Alerts and team management are not available to the assistant.
Security and privacy
- A key sees what you see. It has the same read access to your workspace as you do in the dashboard. Treat it like a password.
- Revoke it any time under Settings → API keys. It stops working immediately.
- It sits in your assistant's config file in plain text. Use a separate key per tool, so you can revoke one without breaking the others.
- Your AI provider sees the data you ask about. Reports and recent events (page paths, UTM tags, device and location summaries, any custom event properties you send) are returned to the assistant, so they are covered by that provider's terms. Mitr itself identifies visitors only by hashed identifiers; see how we count visitors without cookies.
- Visitor data is marked as untrusted. Page paths, referrers, UTM tags and event properties are written by your visitors, so anyone can put text in them, including text aimed at an AI. Reports and recent events are cleaned of hidden characters and returned inside an
untrusted_dataenvelope telling the assistant to treat them as data, never instructions. - No secrets in the conversation.
add_appdoesn't return the new app's SDK secret key, and the webhook tools don't return signing secrets; copy them from the dashboard. - The key is only ever sent over HTTPS (plain
http://is accepted only forlocalhost).
Plans
During early access the MCP server works on every Mitr plan, including Free. After early access ends it is part of the paid plans; Free workspaces then get a message pointing to pricing, and upgrading restores access with the same key.
Troubleshooting
Hosted: open https://mcp.mitranalytics.dev/ in a browser; it should show the server version. Local: run npx -y @mitr-analytics/mcp --version in a terminal. If either works, the problem is in the client config.
| You see | What to do |
|---|---|
MITR_API_KEY is not set | The env block is missing or in the wrong place in the client config. |
| "Your Mitr API key was rejected" | The key expired, was revoked or was copied wrongly, or you're no longer in its workspace. Create a new one in Settings → API keys. |
"missing the 'apps:write' permission" (or funnels:write, qr:write, webhooks:write, delete) | The key doesn't have that permission. Create one with it ticked to use the tool. |
| "early-access period has ended for the Free plan" | Upgrade the workspace; the same key starts working again. |
| 401 from the hosted endpoint | The Authorization: Bearer mitr_live_… header is missing or malformed in the client config. |
| "Mitr is rate-limiting this key" | Too many requests in a short time. Wait a minute. |
npx: command not found | Install Node.js 18+, then restart the assistant so it picks up the new PATH. |
| "Could not reach Mitr" | Check your connection or proxy. Only set MITR_API_BASE_URL if you were told to. |
Configuration
| Variable | Required | Default |
|---|---|---|
MITR_API_KEY | yes | none |
MITR_API_BASE_URL | no | https://api.mitranalytics.dev |
Development
npm install
npm test # builds, then unit tests + an end-to-end stdio test against a stub API
npx @modelcontextprotocol/inspector node dist/index.js # try the tools by hand (set MITR_API_KEY first)
Questions or problems: [email protected]
License
MIT
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.