drafty
Amplitude logo

Build an Amplitude retention dashboard with Claude

Connect the Amplitude MCP server to Claude, ask for a retention and funnel dashboard from your live product data, and publish it to a link your team comments on directly — no BI export, no screenshots pasted into Slack.

What you'll build
A self-contained product dashboard — Day-1/Day-7/Day-30 retention, the activation funnel step by step, weekly active users, and the features that drive return visits — generated by Claude from your real Amplitude data, then published to a drafty.im/canvas/… link. Your team clicks the exact funnel step or retention curve they want changed and leaves a note. Claude reads the comments and ships a revised version to the same URL.

This is an end-to-end example: connect a data source over MCP, generate a dashboard from live numbers, and close the review loop on one link. Total time, start to shared link, is under fifteen minutes. The same shape works for any of the other examples — only the connection step changes.

Here's the finished dashboard, published to a canvas — click any tile or number to leave a comment, exactly as your team would:

Live canvas — comment on any elementOpen ↗

The three moving parts

  1. The Amplitude MCP server gives Claude read access to your Amplitude project — charts, funnels, retention, cohorts, events — through a controlled set of tools. It inherits your own Amplitude permissions; you approve what it can touch.
  2. Claude pulls the numbers and writes a single self-contained HTML dashboard. You iterate on it in the artifact panel until it's right.
  3. Drafty turns that HTML into a stable link your team reviews. Comments pin to the exact element; Claude ships the fix to the same URL.

The generation step is fast now. The part this example is really about is the third one — getting the dashboard in front of people without losing their feedback to a screenshot circled in Preview.

Step 1 — Connect the Amplitude MCP server

Amplitude runs an official remote MCP server at https://mcp.amplitude.com/mcp. You connect once; it authenticates over OAuth, so no key is pasted into a config file.

In Claude Code:

claude
claude mcp add -t http -s user Amplitude "https://mcp.amplitude.com/mcp"

Then run /mcp inside Claude Code and follow the Amplitude OAuth prompt to authorize the account. Confirm it's working by asking "What Amplitude projects can I access?" — a list of your projects means you're connected.

In Claude Desktop: open Settings → Connectors → Add custom connector, name it Amplitude, paste https://mcp.amplitude.com/mcp, and authorize with OAuth the same way. (EU-residency customers use https://mcp.eu.amplitude.com/mcp instead.)

Safety first
The MCP inherits the Amplitude permissions of whoever authorizes it, governed by two role actions: USE_MCP_READ (read) and USE_MCP_WRITE (write). For a reporting dashboard, an admin should grant USE_MCP_READ only — it reads charts and queries data, and has no reason to create or edit objects. Never grant write access for a read-only reporting task.

Step 2 — Pull the numbers

Ask Claude in plain language. It uses the MCP server's read tools to query your charts, funnels, and retention reports directly:

claude
Using the Amplitude MCP server, pull everything we need for a product-health dashboard: Day-1, Day-7, and Day-30 new-user retention for the last cohort; the signup → activation → first-key-action funnel with conversion at each step; weekly active users for the last 8 weeks; and the top 5 events correlated with Day-7 retention. Summarize the figures before you build anything.

Claude calls Amplitude, returns the figures, and you sanity-check them against your Amplitude project before going further. This is the moment to catch a wrong assumption — the wrong cohort, an event you didn't expect, a funnel ordered differently than the report — while it's cheap.

Step 3 — Build the dashboard

Once the numbers look right, ask for the artifact:

claude
Build a single self-contained HTML dashboard from those figures. A retention-curve chart (D1/D7/D30) as the hero, then a step-by-step activation funnel showing conversion and drop-off at each stage, a weekly-active-users trend, and a small table of the features most correlated with return visits. Clean, no external dependencies — inline the CSS and any chart code.

Claude renders it live in the artifact panel. Iterate in place — you're not regenerating from scratch:

Step 4 — Publish to Drafty for review

A Claude artifact link is a preview, not a stable URL — iterate the artifact and the link you already sent now shows the old version. Ask Claude to publish it to a Drafty canvas instead, so the link you share always stays current:

claude
Publish this dashboard to Drafty as a canvas and give me the shareable link.

Claude pushes the dashboard and hands back a drafty.im/canvas/… link that renders on any device. Send it — your team opens it in a browser, no login and no Claude account needed.

Step 5 — The review loop

This is the part that's not obvious until you've done it once.

A reviewer clicks the specific funnel step, retention curve, or number they want changed and leaves a pinned comment — "this activation step looks too high, are we counting users who never finished onboarding?" The comment is anchored to that element, not floating in a Slack thread. Claude reads the comments through the CLI, reruns the relevant Amplitude query if needed, and pushes a revised dashboard to the same URL. The reviewer refreshes and sees the change; the thread stays attached to the element.

The mechanic matters because of what it removes. A Slack message about a chart produces "the funnel number looks off." A pinned comment on the actual step produces "this — define activation as completing onboarding, not just opening the app." One of those produces a correct revision; the other produces a guess.

Keeping it fresh

An MCP-generated dashboard is a snapshot — it holds the numbers Claude pulled when it built it; it doesn't re-query Amplitude when someone opens the link. For a weekly review or a board-ready snapshot, that's fine.

To make it a live canvas that always shows today's figures, copy this prompt — Claude sets up the refresh for you and schedules it to run on its own:

claude
Turn this Amplitude dashboard into a live canvas: every morning, re-pull the latest retention, funnel, and active-user numbers from Amplitude via the MCP server, rebuild the dashboard, and push a new version to the same canvas URL so the link always shows today's figures. Schedule it to run daily on its own.

The link stays stable while the content updates underneath it — see keeping a canvas updated automatically.

What to watch for

Amplitude dashboard with Claude — FAQ

Do I need to paste an Amplitude API key anywhere?
No. The remote Amplitude MCP server at mcp.amplitude.com/mcp authenticates over OAuth, so you authorize the account through a consent screen instead of pasting a key. The connection inherits your own Amplitude permissions — it can only see projects and data you can already see.
Is the dashboard live or a snapshot?
A snapshot. It contains the numbers Claude pulled when it built the file; it does not re-query Amplitude when someone opens the link. To refresh it, ask Claude to repull and re-push to the same URL — or put that on a daily schedule so the stable link always shows current numbers.
Can my team comment without an Amplitude or Claude account?
Yes. The dashboard is published to a Drafty canvas link that renders in any browser. Reviewers click the exact element they want changed and leave a pinned comment with no login required. Only the person connecting Amplitude needs access to the project.
Is it safe to give Claude access to my Amplitude data?
Connect with OAuth and have an admin grant the read-only USE_MCP_READ action — a retention dashboard never needs write. The MCP inherits your existing Amplitude permissions and grants no extra access, every tool call is mediated by the server, and admins can restrict or disable MCP per-project or org-wide.
How is this different from Amplitude's own charts and dashboards?
Amplitude's native dashboards query live data against your tracking plan — the right choice for governed, always-on reporting. This approach is for a fast, shareable snapshot you can spin up in minutes and iterate by talking to Claude, then collect feedback on inline. Different jobs: one is a standing system, the other is a quick reviewable deliverable.