Home (Today)

Overview

Home is where you land after signing in, and it is built to be read in the order it appears. Each role product you own runs 24 x 7 x 365, and they report here. You get what needs your approval, then a short ranked list of what moves revenue today, then one sentence from each role.

The page is deliberately short. Nine roles running checks around the clock produce far more than anyone can act on, so the line is capped at what fits before your first call. Everything a role found but did not raise stays in that role's console, on the Runs tab, and the standup tells you how much each one held back.

What reaches the line is scoped to the roles you hold. A founder on the Founder plan holds all nine; a buyer with two sees those two, and the standup and the line agree about which. An account holding nothing still gets one role's checks free, so the page has something on it before you buy anything. The one deliberate exception is the sample book a new account is seeded with: those come from every role so the page is not empty on day one, and each says sample on the row with a button to swap in your real data.

Your agents run on our own servers on a schedule. There is nothing to install and no desktop app. You connect a data source once, the scheduled passes do the work, and you review what comes back. Nothing is ever sent without your click.

What's on the page

• Waiting on you. Commitments, not suggestions: drafts held for send, and decisions only you can make. This block is never shortened and every row carries how long it has been waiting, because an approval that quietly piles up is worse than no draft at all. Expand a draft to edit it inline, then Copy it, save your revision, or dismiss it. When you dismiss, a one-tap chooser asks why (not relevant, wrong facts, off tone, or just dismiss); the reason is captured so the role's craft library can learn, and "just dismiss" records nothing. Each draft shows a Why this draft panel naming the skill that wrote it, the sources it cited, and any claim it made without one. Approve, Save edit, Copy, Snooze and Dismiss are the same buttons with the same labels here, in the send queue and in every role console: one gesture to learn, and one sentence back when something fails that always says whether anything went out.
• The line. The ranked work, capped, with a rough time cost for the whole list. Every item names the role that raised it, gives one sentence of evidence, and carries a single verb you already use: Call, Send, Book, Draft, Decide or Read. The order is computed, not guessed, and it is deliberately something you can argue with: items rank on how close they sit to a live conversation, whether the check called them today's problem, how long they have waited, and how quickly they clear. Deal size nudges the order where you have deal sizes and never decides it, so the ranking works just as well before you have a CRM.
• The standup. One row per role you own, including the ones with nothing to say. "Ran and found nothing" is a real report and worth reading, and it is what makes the days with something to say stand out. Each row shows its next scheduled pass, how much it held back, and two ways in: that role's console, and its panels.
• Close the day. What your roles raised that did not earn a place today, and a link to the full record of every run.

If a part of your day cannot be loaded, the page says which part failed rather than showing you an empty block. An empty block would read as "nothing to do", and that is the one thing this page must never say wrongly.

Have it sent to you

At the bottom of Home, pick a channel and an hour and your line arrives there at that hour every day. Email goes to your account address unless you name another. Slack posts to an incoming webhook you paste in, which needs no app install and no approval from anyone.

A quiet day is not sent. On the page, "nothing needs you" is a real report worth reading; in your inbox, a daily message that usually says nothing is how you learn to ignore the one that says something. So a message always means there is something.

The message carries the same line, in the same order, from the same renderer the page uses. It cannot disagree with what you see here. Nothing in a delivery can send anything to a customer: it is a message about your day, and every outbound still waits for your approval in the send queue.

What the page remembers

Every item that reaches your line is recorded: what it was, which role raised it, where it landed, and how high. Items you never touch are recorded too, and those matter most. An item you passed over at rank two is the clearest evidence the order was wrong, and it is the signal the ranking learns from.

Skip says "not today" and is worth using rather than ignoring a row. It costs one click, it does not delete anything, and it teaches the order faster than silence does. Dismiss on a draft asks why, and that reason reaches the role's craft library.

Nothing on this page can send. The primary button on a held draft takes you to the send queue, where the send layer's own switch and daily budget apply. That separation is deliberate: one place decides what leaves the building.

Jump anywhere: the Cmd-K palette

Press Cmd+K (Mac) or Ctrl+K (Windows), or click Search in the top bar, on any signed-in page. Type a few letters to jump to a page, one of your pending drafts by title, or a quick action (docs for the page you are on, support, service status). Arrow keys move, Enter goes, Esc closes. Picking a draft lands you on Home with that draft opened and highlighted.

Everything is in the line

Nine kinds of work compete for the same few places: what your roles found, drafts they wrote, decisions the org needs from you, your setup steps, queued outreach, questions worth asking on a live deal, accounts behaving like they are in market, and what the scheduled checks on your book turned up (a deal stalled in security review, a sequence that stopped, a cadence step now overdue). There is no second queue on another page that you are also supposed to be checking.

That last kind used to arrive as a separate email. A weekday note called the Morning Brief carried the top three of them at a fixed hour, to whoever had found the opt-in buried in Settings, and its main button opened a page that had been retired. It is gone. Those checks now go where the rest of your line goes, and answering one here closes it rather than leaving it to come back tomorrow.

Each of those still has a detail page, and the line links you to it. Queued outreach is read and sent from Activity, because sending is a decision the line deliberately cannot make for you. The full intent list, with the signals behind each score, is on Intent. What changed is that you no longer have to remember to open either one.

The line spreads across kinds as well as across roles. If you have forty queued drafts and a busy intent feed, you still will not get a line that is five of the same thing, because a second item from the same source is worth slightly less than the first.

When an account is off limits, the line says so. The outreach rules track who you have contacted and how recently. If an item points at someone you touched a few days ago who has not replied, the row carries a short warning saying to hold off. It is still shown, because you may know something the rules do not; it is never quietly removed and never quietly offered.

What came of it

For a week after you act on something, Home watches for three things and reports what it sees: a real reply from that person or their company, a meeting on your calendar with them, or the deal moving stage. That block sits under Close the day and only appears once you have acted on something.

It never asks. There is no "did this work?" prompt anywhere, because the answer you would get is whoever bothers to reply. Everything reported is read from your own connected mail, calendar and CRM records. An out-of-office does not count as a reply.

Nothing seen is not the same as nothing happened. Items still inside the window with nothing observed are counted separately and are never shown as failures. Work you did outside the product, a reply on a channel we are not connected to, and a genuine dead end all look identical from here, and pretending otherwise would be the easiest lie for this page to tell.

Seven days, then it stops looking. A reply three weeks after a call is a real thing and a poor claim about why it happened.

The order learns from you

The line is ordered by four things: how close an item is to a real conversation, whether it is today's problem or this week's, how long it has been waiting, and how quick it is to clear. There is no model deciding the order. You can read it, and you can disagree with it.

Once you have used it for a few weeks, a daily pass compares the items you acted on against the ones you passed over and shifts how much each of those four counts, for your account only. An item that visibly led somewhere counts for more than one you merely opened, and one where nothing was observed still counts fully on the side you put it: absent evidence is never read as bad evidence. Being shown something and doing nothing counts as passing over it, which is the whole reason every item you are offered is recorded and not just the ones you touch. Snoozing counts as neither, because it says the timing was wrong, not the ordering.

When it moves, Home says so and names what changed, under The order learned something. Nothing about it is hidden, and nothing it does is a decision: it changes what comes first, never what happens. A weight can only drift so far from where it started, so a strange fortnight cannot turn your line upside down.

It needs real use before it does anything, several weeks of use with both items you acted on and items you passed over. Until then you get the ordering everyone starts with, which is not a placeholder.

When you hire someone

On the Founder plan you can hand a role to a teammate from Team admin. They run it: their console, their drafts, their schedule. You stop holding it personally, and your plan does not change price.

Your standup keeps a row for it, and that row changes what it does. Instead of asking you to act on salescron, it tells you how salescron is going: that it ran, how many passes, and when the last one was. That is the whole point of delegating, and it is the difference between doing the job and owning the outcome.

You see that it ran, never what they wrote. The row reports counts and times only, and that limit is enforced by the database rather than by us remembering: the account that reads it has no access to any table holding a draft, a message or a note.

Reversible from the same place, and from the card on Home. If they leave your team, the role comes back to you automatically rather than leaving with them.

Want the role back for yourself and for them? That is a second seat of that role at its listed price. Delegating is free; two people running the same role is two seats.

If you hold every role

On the Founder plan the Work tab holds one entry, Workspaces, and it brings you back to the standup. Eight equal tabs is the problem this page was built to answer, so the menu stops asking you to pick and the standup does the ranking instead. Each role card there links straight into that role's panels.

Nothing is hidden. Every workspace keeps its own address, its own in-page tabs, and its place in the Cmd-K palette, so Cmd+K then a few letters still goes anywhere in one step. What changed is the menu, which is where you choose; the palette is where you go when you already know.

A seat buyer sees no difference. Work already shows only the workspaces you own, and it collapses only for an account holding all nine roles.

Before you buy anything: the free standup

Home works with no subscription. Every day the deterministic detectors read the records you have connected and rank what looks like it needs a decision, and the result is the same page a paying customer sees: Waiting on you, The line, The standup, Close the day. There is no trial clock on it and nothing expires.

What the free tier does not do is write. Drafting a reply, coaching a call, preparing a meeting, all the work a role does with words, runs on a paid seat. So the free standup can tell you a deal has gone quiet and rank it against everything else in the line; it cannot hand you the message to send. The standup block says so plainly and links to the plans.

Two setup steps apply to you: tell your roles about the business, and connect a data source. The other three on the full checklist (adding your AI provider key, choosing where output goes, turning the runtime on) all act on a seat, so they are left off yours rather than sitting there permanently incomplete. Detectors run on your account whether or not you ever open a console.

What raises the ceiling is data, not effort. The detectors read what you connect, so an account with one imported book of business gets a real line and an empty account gets an honest empty one.

Your number

Under the greeting sits one band: your target for the quarter, what has closed, what you marked commit, what is open, and the coverage between them. It appears once you have a target or an opportunity, so a new account is not shown a row of zeroes.

It is not a forecast. Nothing is predicted, weighted by stage, or adjusted by a model. Won is what closed, committed is what you marked commit, and open pipeline is the rest, each a sum of amounts on opportunities you already store. An opportunity with no readable amount is counted out of every total and named, because a silent exclusion is how a total quietly stops being the total.

Coverage is measured against the gap, not the target. Coverage on the full target ignores what is already banked, which makes a nearly finished quarter look thin. Once the target is met there is no gap and no coverage, and the band says so.

A forecast category is read off the opportunity, never inferred from its stage. A stage-to-category mapping is a guess about your process, and this number's whole claim is that it guesses nothing. Set or clear the target from the band itself; the figure lives on this screen, so the field that changes it does too.

Your first login

On day one your roles have nothing to work with, so setup is the line. Instead of a banner over an empty page, the things standing between you and a working line appear as ranked work with the same shape as everything else: tell your roles about the business, add your AI provider key, connect a data source, choose where output goes, and turn the runtime on. A rail at the top shows how many are done. Each one opens the page where that step is actually finished, which for the three console steps is the pane holding the control rather than the console's front door.

You only do those once, not once per role. A founder holding several seats is still one person, so turning the runtime on for any seat counts, and so does choosing one destination.

The rail shows how many are done, and every role card says what that role will do once it can run and how often. That is read from the role's own primary skill and cadence rather than a sentence written about it, so it stays true when a role changes.

While you are setting up, each card names the one thing actually blocking it: your next incomplete step. Roles do not claim their own private blocker, because nothing in the product knows what data each one needs, and a plausible guess on your first screen is still a guess.

Every item that names a company carries a link into that company's record, so the line is a way into the business rather than a list of tasks.

Under the rail, while it is showing, there is a link to the free GTM audit. It reads your website and hands back a company read, an ICP, a competitor list and a sample post in your voice, with nothing connected, on the same engine your roles run on. The rail and the link both disappear once every step is done.

You do not have to go looking for that. A new account is given the sample book at signup and every role's checks are run over it straight away, so the line has real findings on it the first time you open Home rather than an empty page and a promise. It costs nothing and needs no AI plan.

Every row from it says "Sample". A bar above the line says how much of it is sample data and carries one button that takes you to connecting your own. Your records replace the samples and the samples are removed. Nothing computed from the sample book is ever presented as your own pipeline, which matters twice: you would catch it, and the part of Home that learns from what you act on would be learning from a book that does not exist.

Each role's console also offers the same book on a button, to load or clear at any time.

The guided setup at /onboarding is three steps: add your AI provider key, connect your data, choose where output goes. The business question is not among them; it is the screen before, and a new account lands on it straight from signup. Under Choose where output goes the in-app queue is on by default and Deliver to Google Drive toggles there, one switch per owned product (the same switch as the console's Settings tab; it needs Google connected under Connect your data above it). Real Gmail and Outlook mailbox drafts are coming later, after the Google security review, and nothing in the product offers them as a destination in the meantime.

Related

• Persona Console: run and tune one product (schedule, drafts, runs, and settings).
• Activity: the full run history and what each pass produced.
• Receipts: what got written, the data it used, and why.