Support

Sign in with Google or Apple at matters.blackdogranch.com — the first sign-in creates your account, free, no card asked. The same account works on every device.

Action holds what you owe — the next move is yours. Waiting holds what others owe you — someone else moves next, and it stays visible so it can't quietly slip. Notes holds things worth knowing that need no action.

Anything due today or earlier rises to the top of the list on its own — there is no separate Today list to maintain. Priority is meant to stay at medium; raise it only when something truly deviates, so the urgent rows are the ones that stand out.

Open People, tap Invite someone, and send them the message it gives you — however you like: a text, an email, or hold the QR code up to their camera. They tap the link, sign in, and you are connected. The invite works once and lasts seven days.

The name you chose in People, and nothing else. Not your email address, not your list, not any task you have not deliberately sent them. Being connected lets them send you a task and see what became of the ones they sent — that is the whole of it.

Open the task and choose Send to… — on the web it is at the bottom beside Delete, on the phone it sits with Mark done and Delete. Pick the person, add a line saying why if it helps, and send. It waits on their Updates screen until they accept it.

It stays on your list and moves to Waiting, which is what that section is for: someone else moves next. The row tells you where it got to — sent, accepted, or handed back. If you did not want to keep track of it, delete it; nothing is lost on their side.

Your task moves back into Actions and you are told, because the job is yours again. Nothing is recorded against them and nothing changes on their list.

You are asked rather than told. Their copy closes on their list and yours stays open until you agree — because “they say it is done” and “it is done” are different facts, and the task is on your list because you are the one who needs it finished.

While it is still waiting for them, yes — the invite can be withdrawn and the task is simply yours again. Once they have accepted it, it is theirs: you can stop tracking your own copy, but nothing reaches into their list to remove it.

They need an account, which is free — the website works on any phone or computer, and the iPhone app is optional. Tapping your invite link walks them through it.

No. Your AI can propose things onto your own list and nothing else; every send is something you did deliberately. Reaching into another person's list is not a thing an automated run is allowed to do.

Open People and remove them. It ends for both of you straight away, and either of you can do it. Tasks already accepted stay where they are — they belong to whoever holds them now.

Ten tasks can be waiting on any one person at a time. It is there so a connection cannot bury somebody's review screen; work through them and the limit lifts itself.

Open Connections — on the website, or in the iPhone or Mac app — and create a new connection. You get a private URL you paste into your assistant's connector settings (Claude and ChatGPT both speak MCP, the Model Context Protocol). From then on your assistant can read your list and propose changes. On the iPhone there is a step-by-step: Updates, then Connect Claude with MCP.

Open Customize, then Connectors, and use the + beside Connectors to add a custom connector. Name it What Matters and paste your connection URL into the server URL field — the advanced OAuth boxes stay empty, because the token is already in that URL. On a Team or Enterprise plan the owner adds it in Organization settings instead. Claude's own menus are theirs to change, so if the wording has moved, look for custom connectors.

The first one, Connection URL (MCP) — it carries your token and is all Claude needs. The second is the same address without the token and the third is the token on its own, for clients that ask for the two separately. All three appear once, when you create the connection, and cannot be shown again: if you lose them, revoke and make another.

For scheduled runs, yes — choose Allow always on all six. A run that happens overnight has nobody to answer a permission prompt, so a tool left un-approved is simply a run that did nothing. In a live conversation you can approve as you go, but there is little point: the tools are the connection.

No. Your list lives on our servers and Claude's scheduled tasks run in Anthropic's cloud, so neither the phone nor the Mac app needs to be open — a nightly pass lands in Updates while everything you own is switched off.

Say "review my tasks in What Matters" — the answer comes from your actual list, which is the quickest proof it worked. Then paste a meeting transcript or a mail thread and ask for the action items to go in. That is what it is best at, and it is worth doing once by hand before putting it on a schedule.

In a live conversation it can add and complete items for you. On its own — overnight, for example — it can only propose: every suggestion arrives with a reason, and nothing touches your list until you accept it. Turn the same suggestion down three times and it proposes a rule so you stop seeing it, enforced on the server.

Connections lists every token you have created, each named and separately revocable. Revoking takes effect immediately — the assistant simply loses access, and nothing else changes.

Copy a message — an email, a Teams thread, a note to yourself — then tap Smart Add beside the + on your list. It reads what you copied and fills the task in for you: a title you can act on, the details, who owes it or is owed it, a real due date, whether it belongs in action, waiting or note, and how much it matters. You look it over and press Done. Nothing is saved until you do.

A hundred a month on Pro and Pro+, resetting on the 1st. They do not come out of your AI credits — a briefing spends a credit, a Smart Add does not. The free plan does not include Smart Add; pasting still works there, it just puts the text in the title and leaves the rest to you.

Nothing is blocked. The paste still becomes a task — the text goes in the title, and the form tells you in one line that the allowance is spent and when it comes back. The same thing happens if you are offline or the model cannot be reached: you always get the task, sometimes without the help.

No. The button is there whether or not you have copied anything, and nothing is read until you tap it — that is the moment iPhone asks whether to allow the paste, and you can say no. If the clipboard is empty it simply tells you to copy something first and stops. It never looks while the app is sitting in your pocket.

It can, which is why it never saves on its own. What comes back is a draft on the entry form — change any field, or press Cancel and nothing happens. It is the same rule the AI review follows: the model proposes, you commit.

A batch of changes your AI agent thinks your list needs — things to add, things it reckons you have finished, dates that have slipped. None of it has happened yet. Every line is a suggestion, and your list is untouched until you accept. They arrive on the Updates tab, and the number beside it is how many are waiting.

No. What Matters never generates one. You hand your AI agent a prompt and a schedule — the prompt on the Connections page is written for a nightly run — and it goes off on its own clock, reads your list, looks through whatever channels you have connected it to, and files what it found. The schedule lives with your assistant, not with us.

Almost always because nothing is scheduled to run. An empty tab is not a quiet week — it means no agent has been asked to look. Copy the nightly prompt from Connections into your assistant, give it a schedule, and updates start arriving; the Connect Claude with MCP button on that empty tab walks through the whole thing. It is also empty, correctly, right after you have accepted or rejected the last one.

Through a connection you create on the Connections page and paste into your assistant — the same door the briefing uses. It is scoped to your account, it is the only way in, and deleting it cuts the agent off immediately. Any assistant that speaks MCP can use it.

They are the kind of change each line would make. Add is a new item. Done ticks something off that you have already finished. Change edits an item you have — a date that moved, a section that is now wrong. Remove deletes one. Rule is the odd one out: it changes nothing on your list, only what your agent may suggest in future, and lives on the Rules page once accepted.

The title of a proposed new item is editable in place — fix the wording, then accept, and what applies is what is on the screen rather than what the model wrote. The other kinds are accept-or-reject as they stand.

Nothing is sent while you are reading — rejecting a line only strikes it out, and you can change your mind all the way down. One button submits the lot: it applies what you kept and records what you struck out, in the same press. Turning every line down is still worth submitting, because the rejections are the point.

No, and this is worth knowing. Later leaves the whole update waiting and decides nothing — including forgetting the lines you struck out, because none of them were ever sent. Come back and the update is exactly as it arrived.

Because they were about items that are no longer there — you deleted or completed them since the run. They are shown as skipped rather than dropped in silence, so the count still matches what will actually happen when you accept.

A scheduled run cannot: proposing into the Updates tab is the only write it is allowed. Two things do skip the queue on purpose. When you are talking to your assistant directly it acts on what you asked, there and then. And it may push a single genuinely urgent item — someone blocked on you right now — straight onto your list and to your phone. Everything else waits for you.

No. Every rejection is recorded and read by the next run, and a dismissal is treated as an answer. This is why rejecting beats leaving an update sitting — an ignored suggestion teaches nothing. Turn down the same kind of thing repeatedly and your agent will offer you a standing rule, which filters that whole category on the server before it ever reaches you.

No. Each run replaces whatever is still waiting rather than stacking up behind it, so you are never working through a backlog — what you are looking at is always the latest reading of your list.

Once a day a real model reads your whole open list and tells you what you would otherwise miss: what is critical today, who has gone quiet and is worth chasing, what is safe to drop, and any patterns it noticed. It does not change anything on its own — every finding is something you act on with one tap. It is part of Pro+.

You pick the hour in Settings, and the briefing is also the notification — there is no separate "run at 6, tell me at 8" to reason about. Your timezone comes from the device you signed in on, so you are never asked for it; if you have moved, there is a button in Settings to correct it. Set the hour to Never and no scheduled briefing runs.

Yes. The New briefing button on the Briefing tab runs one on demand and shows what it will cost before you press it. It takes about half a minute, and you can leave the app while it works — it finishes on the server and notifies you when it lands.

One briefing spends one credit, whether it ran on a schedule or you asked for it. Pro+ includes 50 a month, which is more than one a day. Nothing else in the app spends credits — adding items, rules, connections and the review queue are all free of them.

No. Each month's allowance is granted for that month and expires at the end of it, so an unused credit does not accumulate. The Credits tab shows what you have been granted, what you have spent and what is left.

The credit comes back. A credit is reserved while a run is in flight and only actually spent when the run succeeds — a failure releases it, and you can try again.

What's critical is open work it thinks needs to happen now — a deadline that has arrived, or something quietly holding up other work. Who to chase is what you are waiting on from someone else, where the ball has been in their court long enough to be worth a nudge. What to drop is work that looks dead. Other observations is a remark about a pattern rather than a single task, and it lists the tasks it was drawn from.

Never. It reads your list and reports; every card is a suggestion and nothing moves until you press one of its buttons. The one thing that happens without a press is that a briefing you never open stays exactly as it is.

On iPhone and iPad, yes — swipe a briefing card right for its first action and left to ignore it, the same two fates the buttons offer, on the side the buttons sit. The web and the Mac app have buttons only. Note that the review screen for AI updates runs the other way round, because every line there arrives accepted: swipe left to reject, right to take that back.

Do today puts the item at the top of your list by dating it today. Chase pushes the date out a week so it comes back at you then, and Escalate moves a stalled item out of waiting and back into your court. Drop deletes it, through the same six-second undo the list uses. Ignore just means you disagree. Every action applies immediately — there is no Apply button and nothing to confirm.

Acting on a finding hides it and stops the next run raising the same judgement about the same thing for 30 days — otherwise tomorrow's briefing would hand you back everything you dealt with today. All of them are listed under Tune → Already handled, and deleting one there lets the briefing raise it again.

Two things: the short list of what matters most to you, and everything you have already handled. It is where you tell the briefing what to weigh, and where you undo a suppression you did not mean.

Name up to ten things — a person, a project, a customer — and the briefing weighs them more heavily when it decides what to notice and what to lead with. It does not move an item between sections, and it does not make something urgent just for matching: a quiet item about a listed project is still quiet. Findings about something on your list wear a Focus chip so you can see it working.

Only if you ask it to, in conversation. A scheduled or autonomous run can never touch the list — the point of it is that these are your words about what you care about, so a term you did not choose would be worse than no term at all. Every term stays visible and deletable in the app whoever set it.

You get one morning notification, not two. On a day your briefing runs, its push carries the same fresh counts plus the headline finding; on a day it does not, the ordinary daily summary arrives as it always has.

The ten most recent are readable in full under History. Older ones stop being readable but still appear in your credit history, so a briefing from months ago still reads as a briefing on that date that cost one credit.

Everything — the full app on web and Mac, AI connections included, with generous daily limits on AI use. Most people never hit them. The free list holds 25 open tasks at a time; Pro holds 500.

Yes, and we would rather say so than call it unlimited. Free holds 25 open tasks, Pro and Pro+ hold 500. The limit counts what is still OPEN, so completing something always frees its place — you can finish and add as much as you like without ever moving the number. Completed items stay on your list for 30 days so you can look back at what you got done, then clear themselves. Your export always contains everything.

Pro raises the AI-use limits for heavier daily use and takes the list from 25 open tasks to 500. Pro and Pro+ are not for sale at the moment: plans are assigned by your administrator, and a plan follows your account to every device.

In your own account on Cloudflare's infrastructure, encrypted in transit, scoped to your sign-in. No ads, no trackers, no analytics scripts, and your data is never sold — see the privacy policy for the full picture.

Connections → Your Data has a one-click export of everything in your account as a file you keep. The same panel can restore from an export.

On iPhone: Settings → Delete my account, which removes the account and everything in it straight away. On the web there is no button for it yet — email support and we do the same thing by hand. Either way it is a real deletion, not a deactivation.

The web app runs everywhere, including phones. There is a native Mac app — the download button is at the top of this page. An iPhone and iPad app is coming soon; your list will follow you there automatically.

There is nothing to set up — every client reads and writes the same account, so the list is simply the same everywhere. Offline edits on the Mac queue and reconcile when the connection returns.