Onboarding
Setting up a product (an idea, first users, or growing)
On this page
Onboarding sets up a new product in a few minutes and replaces the old setup wizard. You answer one question first: Do you have users yet? An idea, first users, or growing (and whether some already pay). Your answer picks the path. If your work email names your site, Mind reads its homepage once and says what it saw ("a waitlist form, looks like first users; you choose"). It never decides for you. The steps are:
| Path | Steps |
|---|---|
| An idea | Users yet?, Which side is harder, Five questions, Mind's drafts, Meet Mind, Your first week |
| First users | Users yet?, Which side is harder, What you have, Mind's drafts, Meet Mind, Your first week |
| Growing | Users yet?, Which side is harder, What you have, Connect, Mind's drafts, Meet Mind, The check-in |
The step you are on is shown as "Step N of M" with the steps above it; earlier steps are links. Start over (before Mind starts reading) returns to the first question with the same product.
While you set up
- Finish later keeps your place and opens the whole app. Every page then shows a slim bar, "Finish setting up · Continue", and Today shows a Finish setting up card with your step. Finishing removes both. Setup is never required before using the app.
- Until you finish or choose Finish later, the side menu, "Ask Mind or search", Ask, the product switcher and the bell are greyed under "Available after setup", and ⌘K and ⌘J open nothing. Still working: How it works, Finish later, Billing and credits, Keyboard shortcuts and Sign out. A link on a setup step (Connect GitHub, Market › Competitors, Today's questions) still opens, with a bar at the top to take you back.
- You always come back to where you were. Sign in, or open any page, and you land on your current step, unless you chose Finish later.
- Onboarding itself costs no AI credits. Only the two optional checks below do.
The three paths
- An idea: five questions (who has the problem, what they do today, the worst part, how often, why you). Skip any
one: Mind marks it as the first thing to find out. Mind drafts a thesis, who it is for and three beliefs to test, each
labelled no evidence yet and sourced to your answer. You land on Today; Journey starts at stage 1.
- Your site. If Mind guessed a site (from your sign-up, your workspace or your email), the Your site card asks Is this your product's site? and shows what it found on it (pricing, support or help, docs, FAQ, changelog, blog). Yes, read it reads up to 15 pages of it and up to 40 articles of each help, FAQ, support or docs site it links to (at most three; your plan's source limit holds, and anything held back is said). Your drafts can then quote it ("From your site · /pricing") as your own words: a belief still says no evidence yet until customers back it. Not this one forgets it; or give the right address. Say yes after drafting and Draft again with your site appears (your first drafts are kept). No credits; nothing is sent.
- Research my market first (on the last question's screen, on by default, costs no credits, once per product). While you review the drafts, Mind looks on public pages for who else does this, who else buys it, your industry (a profile and its news of the last two weeks), what public sources say about your thesis, who it's for and your riskiest belief, a product that tried this and stopped, and where your people gather. It takes a few minutes; you can go on while it runs (the jobs tray shows it). It searches only short phrases from your drafts, never anything from your customers. Nothing is sent, nothing it finds counts as evidence, and your answers are never changed. Untick it and Mind reads nothing from the web, not even for competitors. See What Mind found below.
- Also check public sources (shown instead once a product has had its free research, on the last question's screen, priced there: 3 credits, refunded if Mind finds nothing it can quote). After drafting, Mind searches public pages (up to 3 searches and 5 pages) on your thesis, who it's for and your riskiest belief. Each draft shows What public sources say: quotes that support it and quotes that cut against it, dated, linked and marked public. They never count as evidence and never change your answers. What the pages raise that you didn't mention becomes a question under Today › Mind asks; products they name show up as competitors to verify.
- First users: add your site, help center, changelog, pricing page, app listing and competitors, and drop up to 10 files (PDF, Word, PowerPoint, Markdown or text, 10 MB each). Each link says what it becomes; each file gets a quick skim ("thesis, segments and pricing spotted"). Press Build: Mind reads what you gave it (15 pages of your site), writes its summary, looks for competitors, and drafts a thesis, segments, the riskiest beliefs and a test. The same Research my market first switch sits above Build (on by default, once): once the drafts are in, Mind adds the rest of the research (its competitor search is the one Build already ran).
- Growing: the same, then the Connect step, then your first check-in places the product on Journey.
The Connect step (growing path)
Two cards, then a catalog.
- GitHub. What shipped, from commit titles on your main branch. Never your code. Connect GitHub opens Settings › Connections in a new tab: turn on Sync, add the webhook it shows, then reload this step. Once connected the card lists the last commits.
- Stripe (totals only). Who pays and how much, from your plans or your one-time payments. Paste a restricted
key (
rk_live_…orrk_test_…, never a secret key) with Read on Charges (one-time payments: sold per use, per order, per event) or on Subscriptions and Products (plans billed every month or year). It is kept encrypted and never shown again; until it is connected nothing about money is guessed. The card then shows counts: paying customers, subscriptions and MRR, or for a product paid per use, customers who paid in 90 days, payments and revenue in 30 days. No customer's id, name or email is kept by totals. Disconnect and forget the key undoes it.- Read my customers too (a product admin's tick above the key): the same key then reads your customers into People, with their plan and payments, and the card says what is kept: per customer the email, first name and last initial, plan and payments; never the full last name, address, phone, tax id or card; "Mind's AI gets a key like C-12, never a name or email from People; words people wrote reach it as written." It also needs Read on Customers, and with plans Invoices and Prices.
- If the key lacks them, totals still connect and the card says Customers not read. Add to the key: with each permission as a chip (Customers: Read). Add them in Stripe and use Read customers too, which opens Settings › Connections › Stripe.
- Once customers are read the card shows how many are in People, Reading customers while it runs, and Manage in Settings. Details: Stripe customer data.
- A key Stripe refuses is said under the field.
- Everything else is listed as Connect it later, or ask for it, by kind (money, usage, what customers say, where the work is, docs). A tool that connects now has Connect (or Import once for Notion and Google Drive) and opens Settings › Connections; Request it counts your ask.
The drafts
When you press Read it and draft, the page shows each thing Mind does as a row with its own live count (your site's pages are read as you leave the step where you confirmed them, so the wait is shorter); you can leave and the bell says when the drafts are ready. A growing product's drafts are asked for in three calls at once (what it believes and who it is for, its product areas, its features), so about twenty come back together.
Drafts are grouped by kind, each kind once as a heading: Your thesis, Who it's for, Beliefs to test, Tests, Product areas, Features and Competitors Mind found. A growing product gets a thesis, three segments, up to eight areas and eight features; first users a thesis, three segments, three beliefs and a test; an idea a thesis, one segment and three beliefs. Each card says what it was drafted from (Why?: the quote, its page or slide) and shows a badge only when it is not plainly sourced: no evidence yet for a belief or test, possible · verify for a competitor.
- Select on a card adds it to the one plan; Accept N selected applies every selected draft as one change you can undo for 24 hours (Undo). While that plan stands, other drafts can't be selected until you undo it.
- Edit changes the title and sentence. ⋯ holds Not now and That's wrong; Put it back returns a draft you set aside.
- Anything not accepted stays a draft ("The rest stay drafts"). You can go on without accepting.
- Mind drafts fewer things rather than guessing. If it had little to go on: Mind found nothing it could back up with a quote, so it drafted nothing rather than guess. Add a deck or a page in Sources and it drafts again. If one of the three calls failed after several tries: Mind drafted part of this: one of its calls failed after several tries. Write the rest yourself, or add more in Sources. If all failed: Mind couldn't draft just now, after several tries. Write your own, or add more in Sources.
Once you accept a thesis, Look for competitors (2 credits, refunded if it finds none) searches public pages from your thesis and who it's for and lists what it finds as possible · verify; nothing is watched until you choose. It is on the drafts step and on an empty Market › Competitors, and the thesis then says Others doing this: N found.
What Mind found (the first research)
Beside the drafts, What Mind found on public pages opens with one line (public pages, each sourced and dated, never counted as evidence) and a strip of the parts, each Looking, Found or Nothing found; tap one to jump to it. Parts fill in as each lands:
- Who else does this: competitor cards (up to 30, the first 10 shown, Show more, See all N goes to Market › Competitors) with Watch or Not now.
- Who else buys this: kinds of customer you didn't list, found on your competitors' own "who we serve" pages and on pages about your category. Each card has the kind, why they buy, the page's own quote, and where and when it was read. Add as a customer makes it a candidate segment beside your main one (on Market › Positioning and Learn › Compare) with Undo; Not for us sets it aside, Put it back brings it back. A kind you already listed, however you wrote it, is never shown. The Who it's for draft says Mind found N more who buy this.
- Your industry profile (Confirm, Edit on Market › Industry, or Not now) with your industry's news.
- What public sources say on each draft.
- Products that tried this (Keep, Not relevant, or Not now).
- Where your people gather: places as cards with their kind (Community, Forum, Group, Q&A…), what they are and the day they were read; the first 10, then Show more; the searches to try sit below.
Every item names where it came from and when: your answers, your site, or a public page with its date ("Named on news.example.org · 1 Mar 2026"; "From the model's general knowledge, no page behind it"). Watch and Confirm apply at once and can be undone for 24 hours; anything you set aside stays a labelled suggestion on its page (Market › Competitors, Market › Industry, People › Find). A part that found nothing, or that Mind skipped (to stay inside the research's cost limit, or your workspace's AI allowance), says so in its own words.
Meet Mind, and your first week
Before you land, one screen shows what Mind knows about your product, and what it doesn't.
- What Mind knows about {product}: four counts in one card: your answers, your site's pages read, the drafts (and how many you accepted), and what the first research found on public pages. Review and See them go back to the drafts step.
- What was searched on the web is said once, as What's searched (or No web search when none went). Public pages are marked public and never count as your evidence.
- The list of what Mind doesn't know holds only what you can act on, each with one action. "Beliefs with no evidence yet" is what your first week tests, and the step's button reads Your first week.
- When Mind is wrong: three short ways in one row. Say that's wrong, edit or forget a memory, or see where a claim came from.
Your first week (idea and first-users paths) has up to four steps, each a card:
- Find people who have this problem. One LinkedIn search per kind of customer you named, worded the way people write their own headline (a kind and the roles that buy it). Choose People you can reach (your 1st and 2nd connections) or Everyone; Google is another way in; the copy button copies the search. Below it, Where they gather shows the places the first research found (All N in People › Find). Then Then ask for 20 minutes: the email Mind wrote, opened in Gmail, your email app, or copied. You pick who gets it and send it yourself.
- Know what to ask: your riskiest belief, with Write my questions (a script).
- Run one small test: the test drafted for you or, when none was, Draft the test for a belief a small test can settle (a price first). Decide what counts as a yes before it starts.
- Your Journey's stage 1 as a progress card.
MindBacklog never sends anything. Prices beside buttons are hidden for now; charges are unchanged, and an action your balance can't cover still says so.
The first check-in (growing path)
The last step is a card, Four questions place {product} on its Journey, with Start the check-in (see The check-in).
Finish later, and coming back
While setup is unfinished, the app waits for it. The side menu, Ask Mind or search, Ask, the product switcher and the bell stay on screen, greyed, under "Available after setup", and ⌘K and ⌘J open nothing. How it works, Finish later, Billing and credits, Keyboard shortcuts and Sign out still work (on a phone, Sign out is at the foot of the menu). Links on a setup step, such as Connect GitHub or Market › Competitors, open with the menu greyed and a bar at the top to take you back.
You always come back to where you were: sign in, or open any page of the app, and you land on your current setup step, even days later.
Finish later is the way out, and nothing is required before you use the app. After it the whole app is yours: every page shows a slim bar, "Finish setting up · Continue", and Today shows a Finish setting up card with your step. Signing in again doesn't pull you back into setup. Open /onboarding or the bell to continue. Finishing setup removes the bar and the card.
Onboarding itself costs no AI credits, and neither does its first research (once per product); only the two optional checks above do. Credit costs aren't shown beside buttons for now; charges still apply as listed on the AI credits page, and an action your balance can't cover still says so.
FAQ
I clicked away and the menu is greyed. Your setup isn't finished. Use the bar at the top, or Finish later to open the app fully.
Why doesn't the Connect step show on my idea? Only the growing path has it. Connect tools later in Settings › Connections.
Why was a key refused? Use a restricted key, not a secret key, with the permissions the card lists.