mcp server

ask your analytics from your AI

connect claude, chatgpt, cursor or claude code to dashi with one url and a sign-in. ask about your visitors, sources, revenue and goals in plain words — or let your AI add dashi to a new site for you.

https://www.dashimetrics.com/mcp

free, like the rest of dashi. read-only for your numbers; it can't delete sites or data.

add it to your AI

claude

  1. in claude, open settings → connectors → add custom connector
  2. name it dashi, paste the url above and click add
  3. click connect, sign in to dashi and approve

chatgpt

  1. in chatgpt (paid plans), open settings → apps & connectors → advanced settings and turn on developer mode
  2. back in apps & connectors, click create: name it dashi and paste the url above
  3. sign in to dashi and approve, then use dashi in any chat

cursor

  1. click add to cursor — or open cursor settings → mcp → add a server and paste the url above
  2. when cursor asks, sign in to dashi and approve
add to cursor →

claude code

  1. run the command below in your terminal
  2. type /mcp in claude code, pick dashi and sign in
claude mcp add --transport http dashi https://www.dashimetrics.com/mcp

on claude team/enterprise or chatgpt business, a workspace admin may need to add it.

try asking

  • “how did last week compare to the week before?”
  • “which sources bring paying customers?”
  • “why did traffic spike on tuesday?”
  • “where do people get stuck?”
  • “add dashi to my site”

what it can't do

  • ✕ delete sites or analytics data, or change your sharing, privacy or billing settings
  • ✕ see your stripe or search console keys
  • ✕ see globes you're a teammate on

you approve each app on a dashi sign-in screen, and can disconnect it any time from the AI icon on your globe — access stops immediately.

16 tools

read

  • list_sitesthe user's sites on dashi, with domain, timezone, when tracking started, and whether stripe revenue and google search console are connected. call this first; other tools take a site.
  • get_overviewheadline numbers for a site and period, as on the dashboard: visitors, pageviews, bounce rate, average session, conversion rate, online now, returning, revenue (total and verified), errors — plus the trend over time. set compare=true for the previous period too.
  • get_breakdowntop items for one dimension, with visitors (and revenue for acquisition, geography and device dimensions): sources (channel, referrer, campaign, term), pages (page, entry, exit), geography (country, region, city), tech (browser, os, device) or custom events (event).
  • explain_periodwhy a given hour, day or month looked the way it did — the dashboard's 'click a chart bar' story: how it compares with the period average and the bucket before, what drove it (sources, referrers, landing pages, campaigns, keywords, countries), and the revenue in it.
  • get_revenuerevenue for a site and period: totals (and how much is verified), revenue by a dimension (channel, referrer, campaign, country, device…), and recent purchases each with where it came from — the dashboard's 'this sale came from there' story.
  • get_heatmapthe dashboard's heatmap: visitors or revenue for each hour of each day (or weekday, for long periods), in the site's timezone unless you pass one. good for 'when do people visit / buy?'.
  • get_goalsconversion goals for a site: which events count as goals, completions and conversion rate per goal, events that fire but aren't goals yet, recent completions each with who and where they came from, and progress on the globe's target goal (e.g. '1,000 visitors this month').
  • get_funnelsthe site's funnels and how many people reach each step, with drop-off between steps; compare=true adds the previous period.
  • get_site_healthwhat's going wrong on the site: top javascript and server errors (with how many visitors hit them), friction hotspots (pages where people get stuck — rage taps, dead clicks), and how far people scroll on each page.
  • get_ai_crawlerswhich AI bots (OpenAI, Anthropic, Google, Perplexity…) crawled the site, how often, and which pages they fetched. bots are never counted as visitors.
  • get_search_keywordsgoogle search console queries for the site — clicks, impressions, click-through rate, average position, and estimated revenue per keyword — when search console is connected in dashi.
  • get_livewho is on the site right now — each visitor's city, device, browser, where they came from, current page, time on site, visit count and whether they seem stuck — plus the live activity log for the last 30 minutes (visited, completed, paid, exited, stuck).

set up

  • add_siteadd a website to the user's dashi globe so it can be tracked — the same as 'add domain' in the app. adding a domain the user already has is harmless (it returns that site). then call get_setup_instructions.
  • get_setup_instructionsthe exact install steps dashi generates for a site: the tracking script tag (plus a next.js version) and a step-by-step prompt to follow in the user's codebase — optionally with revenue tracking, custom events, server error tracking and an ad-blocker-proof proxy. anything that needs a secret or a stripe key is listed as a manual step in dashi.
  • check_setupis the site set up? whether the tracking script has reported in (and when it last did), whether stripe and search console are connected, which conversion goals are set, and the next step.
  • set_goalsmark which events count as conversions for a site (e.g. signup, checkout, or $purchase for revenue) — the same as the app's conversion goals settings. only adds or removes the names you pass; repeating a call changes nothing.

privacy · claim your globe →