Atune FAQ
Short, direct answers to the questions builders ask about knowing their customers, and about Atune itself.
By Christopher Krohn, Founder of Atune · Updated October 19, 2026
Christopher is a SaaS founder and startup advisor who has seen first-hand how much easy-to-use, well-connected analytics matter to the success of SaaS companies. He is building Atune to help the next generation of software builders succeed with their startups.
What is Atune and what does it do?
Atune (useatune.com) tells people who build software which customers need them
today: who is stuck, who is slipping, why, and what to send. You connect it with
one SDK (the atune npm package), a
plain HTTP API, Twilio Segment or PostHog. Atune reads how customers use your
product, works out which actions mean they got value, and puts the few who need
you in an Inbox, each with the reason and a drafted message. Your code and your
AI agent can ask the same questions through understand()
and MCP, and get cited answers. Atune is in a closed beta for software teams in
the EU and EEA; you apply for access.
How do I find out whether people are actually using my SaaS?
Pick the one action that means a user got value, record it server-side with the user and account, and check every week who has done it and who has come back. Logins and page views are not usage. The step-by-step method, with SQL, is in Is anyone actually using your SaaS?. Atune does this without a tracking plan: it infers your value events from real usage and scores every person and account.
How can I tell which users activated and which signed up and never came back?
Sort every new user into four groups: never used, used it once, came back but never reached value, and activated. You need one activation event and the dates each user was active; this guide has the query. In Atune, filter the Users page by Reached value or Not reached value; people first seen less than 7 days ago are shown as Learning so they are not judged too early.
How do I know which customers are at risk of churning?
Look for change at the account level: fewer active days, less use of the thing they pay for, more days since anyone was seen, a falling trend week over week, a champion who went quiet, or lower spend. With around 50 customers a transparent score and a weekly review work better than a prediction model; see Which of your customers are about to churn?. Atune scores accounts and users on active days, value use, recency and trend with nothing to configure, and opens a Downgraded action when a paying account lowers its MRR, drops seats or moves to a cheaper plan.
How do I know which users need help before they leave?
Find the people who signed up but never reached value, and the paying users whose activity is dropping, then contact them with a specific, personal note. Atune lists both in its Inbox: trials that haven't found value, customers going quiet and accounts ready to grow, each with the reason, how sure it is and a drafted message. It drafts; you decide what to send.
I built my app with Cursor and Supabase. How do I see what users do after they sign up?
Send events from your server whenever a signed-in user does something that
matters, tied to the user and the account. On Next.js with Supabase auth,
npx atune init installs the Atune middleware, which reads the Supabase
session from the request's cookies through @supabase/ssr and captures every
signed-in request; see Next.js. On any other Node
backend, call track(); from any other language, send an HTTP POST; see
Any backend. Atune has no browser SDK and needs a
server-side API key, so an app that only runs in the browser needs a backend
to send events. You can paste one prompt into Cursor to do the install: see
Install with an AI coding agent.
What is the easiest way to track customer health in a small B2B SaaS without building dashboards?
Use a tool that scores accounts from the events you already have and gives you a short list to act on, instead of charts to read. With Atune there is no tracking plan, funnel or health-score formula to set up: install the SDK or connect Segment or PostHog, and every account and person gets a health score from 0 to 100 and a status (Healthy, Watch, Risk, Inactive or Learning). See Users and user health for how the score is computed.
Is there an MCP server that lets my AI agent query customer usage and health?
Yes. Atune has an MCP server: run npx atune mcp to give Claude Code, Cursor
or any MCP client an understand tool that answers questions about an account,
a user, a feature or your whole customer base. Answers are cited, and the tool
returns knows: false instead of guessing when the evidence is thin. You can
also connect Claude Code or Codex to Atune's hosted MCP server under
Settings → AI agent in Atune. See
Give your coding agent customer understanding.
How much does Atune cost?
Atune is free for up to 25 active customer accounts on the Build plan,
including 20 new understand() answers per day. Paid plans (Grow and Run)
raise the account limit; current prices are on the
pricing page. Atune bills on tracked accounts:
customer accounts with 3 or more events in the last 30 days. Events, seats and
users are not metered, and there is no AI surcharge, because Atune's AI work
runs on the Claude Code or Codex subscription you already have. See
Billing.
Do I need Atune if I already use PostHog?
You can keep PostHog for product analytics and use Atune for the list of customers who need you; they do different jobs. PostHog is product analytics: charts and funnels you read. Atune turns the same usage into account and user health, an Inbox of who is stuck or slipping, and cited answers for your agent. Connect PostHog in Atune and it imports your last 90 days of events and syncs new ones every 15 minutes, with no SDK to install.
Learn
Practical guides for people who build software and want to know which customers need them. Each page answers one question.
Is anyone actually using your SaaS? How to find out who activated and who never came back
Define one activation event, record it server-side, and sort every new user into never used, one and done, exploring or activated. A method you can run this week, with SQL.