Settings

Settings opens a list of rows rather than a wall of controls, and the answer beside each name lets you read the whole state before you open anything.

• Account — who you're signed in as, the timezone, your password, what the phone app needs, the calendar link, a download of everything you have, and the way out.
• Accounts — on a LifeOps that holds more than one account: who has looked into yours, and for its owner, which account may look into which. Accounts below covers it.
• AI — one switch for the whole thing, then one per job, each with its model and its prompt.
• Menu bar — which apps stand on the bar, in what order, and which page opens first.
• Notifications — in the phone app only: whether this phone can be reached, which kinds of news reach it, and a test that says what became of one.
• Areas — making, renaming and colouring the areas of your life.
• Appearance — the colours the app is drawn in, and what a goal row shows. The row beside the name says whether either has moved off what your LifeOps shipped with.
• Integrations — the outside services your account runs on and the keys they take, one row per service. A key belongs to your account rather than to an app, which is why this row stays up here.
• Apps — a heading over one row per app that has something to set, in the order the menu bar lists them: Journal, Tasks, Goals, then MCP and Sleep, then any app an extension brought with a settings page. Each row opens that app's page, and Apps below says what is on one.
• Backend — which backend this app is reading and writing through, and how to point it at another one. The admin guide covers it.
• Guides — this manual, one page per screen. It was a ? in the corner of every screen until Sep 2 2026.
• Development — the work on the app itself: what's being built right now, what's queued, what's waiting on you, and the commit the backend and the web last deployed.
• Components — every piece the app is built out of, drawn on its own and in each state it has: the buttons, the fields, the menus, the rows. Everything on that page is made-up filler, so pressing anything there changes nothing of yours. It is there so a piece can be judged in one look, which it never can be on a working screen.

Settings is off the menu on both surfaces, since the menu is for the places you work all day. It is the gear in the top corner, beside the magnifier and the chat bubble, and that corner is the same three icons at every width and in the installed phone app. The corner is there wherever the bar is, so Settings is one tap from wherever you are standing rather than a trip to Day first. In a browser the Config tile on the Apps page opens it too.

Which apps stand on the bar is arranged on the Menu bar page, by dragging, and dragging an app off the bar is how it leaves; there is no switch that turns an app off. Apps and the bar covers it.

The phone app puts its settings on one scrolling screen instead of a list of rows, under a Server row naming the server it is signed in to. Its Apps section is the same list of rows in the same order, after Integrations, and each row opens that app's own screen with the same things on it. The other differences run both ways. Notifications is on the phone and has no row in a browser, since it answers a question about this phone alone. Changing your password, Backend and Development are in a browser and not on the phone. Anything you change on one shows up on the other, since both are asking the same LifeOps.

Account names who this LifeOps belongs to and which zone your days are measured in. Every date the app works out resolves through that zone, so the zone decides which day an entry belongs to and when a day rolls over.

Both apps follow the device you're holding. Land somewhere and your days move with you, with nothing to set — the phone has always done this, and the web does now too. A zone you pick by hand is kept until you land somewhere new. On the web, tapping the zone also opens a search over every zone there is, each showing the current time in it, for when you want a zone that isn't where you're standing.

Changing the password is here too, on the web: the password you use now, the new one, and the new one again. A new password needs at least twelve characters, and every refusal ends by saying your password has not changed. A change signs every browser out, this one included, and sends you to the sign-in page to come back in with the new password. A forgotten password is still a request to whoever runs your LifeOps, covered in the owner's guide under Your login.

Your Google Calendar connects here too, under Calendar: one button, Google's own permission page, and you're back. Once connected, the section lists every calendar on your Google account. A tick beside each chooses which appear on your Day — work, home, and your own, not just one — and a life area beside a ticked calendar files its meetings under that area, so a work calendar's meetings read as work and the time log a Done meeting becomes does too. Your meetings arrive as tasks and appear within fifteen minutes, and Disconnect in the same place forgets the link. If the section says the instance isn't registered with Google yet, that one-time step belongs to whoever runs your LifeOps.

A connection made before the app could answer invitations reads your calendar and cannot decline anything on it, and the Calendar section says so: Google can read the calendar but cannot decline for you. Reconnect to allow it. Reconnect stands beside that line and takes you through Google's permission page again, which is the whole fix. Until you do, Cancel & Decline on a meeting comes back refused, and Cancel task beside it still works, since that one takes the meeting off your day without telling Google anything. A connection that can answer invitations shows no such line.

Your phone on the same screen shows the server address and mints an access token. The phone app does not need either of them any more: it signs in with the username and password you use here. Keep this for a phone pointed at a server its build does not know about, and for handing somebody a way in without handing over a password. Getting in covers both routes.

Your data takes everything in your account out as one zip: every entry, task, note and setting, plus your photos and attachments. A large account takes a few minutes to prepare, and a download that stops early will not unzip, so start it again rather than trusting a short file. Under the button is one line saying whether anything holds an off-site copy of this LifeOps and when it was last made. Bringing data back in is not built, so this is a way out and a way to hold your own copy rather than a way to move between one LifeOps and another.

Sign out asks twice on both apps, and Sign out for good is the press that does it. Stay signed in is the way back out of the question. In a browser it ends the session and returns you to the sign-in page. On the phone it also clears the stored token off the device, so getting back in means signing in again.

A LifeOps can hold more than one account, each with its own journal, tasks and notes. An account opened to yours shows in the pill in the top corner, beside the magnifier, and choosing it shows that account's LifeOps without signing you out. Back to your LifeOps in the same menu returns you.

A read-only account says so twice: the menu row ends read-only, and every page shows a band at the foot reading Viewing and the account's name. You can read everything there and change nothing. A change you try is refused with a sentence saying the account is open to you read-only, and nothing in it moves. An agent that works for you, and the person an agent works for, are open both ways for reading and writing; what you write there is marked as written by you, never as the account's own.

Settings › Accounts lists every visit into your account: who looked, whether read-only, through which app, and from when to when. Requests within half an hour of each other count as one visit.

The owner of the LifeOps decides who may look into whom on the same page. A row names two accounts and picks Read-only or Read and write; Open it saves it, Take back removes it, and right-clicking a row switches its mode. Opening one account to another does not open the other back. The owner sees every account's visits rather than only their own. The web app has the switch; the phone app does not have it yet.

The list of life areas is yours to write. Adding one takes a name and nothing else. Tapping a name lets you rename it, and the new name shows everywhere the old one already was. The whole app files by that list. Tasks group by area, journal entries take one, and the Day shows their colours.

Each area also has a colour, picked from a palette of thirteen. An area you never pick for reads Automatic and takes the next colour no other area is wearing, going down your areas in their order, so a fresh LifeOps is already colourful rather than grey and no two areas share a colour until all thirteen are in use. A colour you pick by hand stays yours, and an Automatic area that was wearing it moves to a free one. Automatic is also the way back if you change your mind.

Each area can also say what it is, in a line under its name that you tap to write. That line is what the AI reads when it files an entry, a task or a note, so an area described in your own words gets filed the way you meant it. On the web the line saves itself a moment after you stop typing, and again if you click away or leave the page, and it says Saving… and then Saved while you write.

The areas stand in the order you give them, and that order is the one every area list and picker shows and the one the AI is given when it files. On the web, drag an area by the handle at its left, or use the up and down arrows that appear at its right when you point at it. With the handle focused, the arrow keys move it too. Moving an area changes only where it stands: its colour, its words and everything filed under it stay as they were.

Under the list is a field for adding one, and under that a line counting how many of your areas have a colour you picked rather than Automatic. There is no delete. Removing an area would unfile every task and every entry filed under it with nothing to undo it, so renaming covers a typo and an area you've finished with simply gets nothing new.

Appearance holds two answers about how the app looks. Each one starts at what your LifeOps shipped with, so nothing you see moves until you change one of them here. The row on the settings list says As shipped until one has moved, and Changed after.

Colours lists every colour the app is drawn in, each with its name, the colour itself, and the value in force. Type a value in the form #1A2B3C and leave the field to store that one colour; the others stay where they are. Reset to defaults puts them all back at once, and it is greyed out while none of them is yours.

Goals is the answer for a goal row, either its counts or the work under it.

Some of what this app does runs on outside services, and those services bill whoever owns the key. Settings → Integrations is where those services and their keys live. Each row is one service: what it is used for, whether your account holds a key for it, and a field to store one. A key is stored once and never shown again. Remove on the row takes it out, and asks first.

Three services take a key. Anthropic is what every AI job and Chat run on. ElevenLabs turns a recording you post into a transcript. Resend emails you when a background job breaks or recovers. Google Calendar is connected instead of keyed, under Account.

A key belongs to your account and pays for your calls alone. Yours runs on that key and on nothing else, and no other account can spend it. There is no instance-wide key and no deployment variable behind the scenes. An account with no key for a service cannot use what that service powers, and the feature's switch, on the app's page under Apps, says so on the row rather than promising something and doing nothing.

Apps is a heading over one row per app that has something to set, in the order the menu bar lists them: Tasks, Goals, then MCP and Sleep, then any app an extension brought with a settings page. A row opens that app's page, headed by the app's name and holding whatever that app lets you set about itself. Tasks holds the task system and the rows that open your statuses, your recurring tasks and your reminders. Goals holds your kinds. MCP holds the assistants that have connected. Sleep holds the morning reading. Journal, Day, Events and Notes have nothing to set, so they have no page and no row; the Journal's one switch, for transcribing a recording, is on the AI page with the other jobs.

There is no switch that turns an app off. Each app's page opened with one for a morning, and it went the same day: an app you do not want in front of you comes off the bar on the Menu bar page, and that is the whole of it. Your work in an app off the bar is untouched, and the app still opens from the Apps page at the end of the bar.

The AI page starts with one master switch, the AI on or off for you, and under it one switch per job. The list of jobs comes from the server itself, so a job cannot run without a switch appearing here. Each row says what its job does and exactly what it sends when it runs. Under that, Model chooses which model the job runs on, and Prompt opens the words it is given, for you to edit and save. A row whose prompt you have edited says so, and Reset to default puts the shipped wording back.

Life area filing files a journal entry, a task or a note that has no life area under one of yours, and it sends the thing's words, your areas with the descriptions you wrote for them, and a few of your own past filings and corrections as examples to Anthropic's API. It runs as background sweeps, about two minutes behind an entry or a task and five behind a note, rather than in the moment you save. An entry it cannot place stays unfiled and is never asked about again; a task or a note it cannot place is asked about again when your list of areas changes. A task under a goal takes its goal's area without asking the model at all. Area descriptions drafts what a life area you have only named means, sending your area names and the descriptions you have already written. The draft shows the word draft after it on the Areas page, and it is yours to edit or clear. Transcribe a recording turns a video or a voice memo you post into a cleaned-up read-through of what you said; it is the one job that sends a file, and it needs an ElevenLabs key on Integrations beside the Anthropic one. What changed in a version writes the line saying what moved between two versions of a note, a goal or an event. Quick Capture, Journal or Task decision decides whether a Quick Capture on the Apps page is a task or a journal entry, and trims a task to its words.

Off, entries save with no area filled in. Everything you type is still stored, still searched, and every control still works. The owner's AI page covers where the text goes once it leaves.

The switches are on the page whatever your account holds, so this page can never lock you out of the settings that would fix it. With no Anthropic key stored, a line under the master switch reads No Anthropic key is stored, so nothing here runs. Add one in Settings → Integrations. Its last three words are the way to Integrations, and storing a key there takes the line away. In a browser, the AI row on the Settings index reads On, no key in that state, which is the switch and the key each answering for themselves.

Settings → Apps → Tasks starts with Goals & task detail, which turns the simple checklist into the full task system, adding statuses, task types, three ways of grouping the list, and a filled-in Goals tab. The same tasks are underneath either way, so turning it off hides the detail and destroys none of it.

Three rows open from that page, and each is a page of its own, indented under Tasks in the rail of a wide window. Your statuses is there once the switch is on. It holds the words a task can stand in, under the four groups the app reads, each group a card that folds, and Your statuses covers it. Deleting a word that is in use asks where its tasks should go before it takes anything away, and a status keeps the order it is in. Recurring tasks and Reminders are there whatever the switch says, and Recurring tasks and Reminders below cover them.

Goals & task detail covers what that switch gives you and how to work it. What a goal can be is on the Goals page, not here.

Settings → Apps → Goals holds, headed Kinds, the list of what a goal can be: one plain list with no groups and no word held back from renaming, in the order the Goals page draws its sections in. Opening a kind lets you rename it, + Add kind under the list adds one to the end, and you move a kind by dragging it up or down. Deleting a kind that is in use asks where its goals should go before it takes anything away. Your kinds covers the list and how each surface works it.

The list is part of the fuller task system. With Goals & task detail off, the section says so and names the switch under Tasks instead of drawing an empty list.

Settings → Apps → Sleep holds one switch for the morning sleep reading, off if you don't want the entry at all (the readings already in your journal stay), and the life area each morning's entry is filed under. Filing it under no area is a real answer, and the entry is still written. Delete the area you chose and the app files the nights under no area again.

Notifications in the phone app starts with This phone. Turning it on is where the app asks your phone for permission, unless setting a reminder has already asked for it. Under it is a switch for each kind of news the app can send. Every kind starts off, and switching one on is what lets that kind through. Reminders covers the nudges you set, and Background jobs tells you when something that runs on its own stops working, and again when it starts working. A LifeOps run with notifications off altogether says so above the switches, rather than storing answers nothing will honour.

Send a test sends one notification to this phone now and says underneath what became of it. There are four answers and it rounds none of them into another. Sent, not confirmed yet means the message left and Apple has not answered for it yet. The screen keeps asking for about a minute, then says Not confirmed rather than guessing. Arrived appears only once Apple has confirmed the delivery. Didn't arrive comes with Apple's own reason. And when nothing left at all, the line names which nothing it was, since no phone registered, notifications off on this server, and a notification service out of reach are three different problems with three different fixes. The test ignores the kind switches, so it answers whether this phone can be reached even with every kind off.

Recent keeps the answer after you have stopped looking. The last ten notifications the app sent are listed with what became of each, so a nudge that never turned up is something you can go back and read the reason for.

Turning This phone off stops the app reaching this device and leaves your phone's own permission alone, since only you can give that back. The kind switches keep their answers while it is off, so turning the phone back on does not ask you everything again.

Reminders, a row on the Tasks page under Apps, opens on what you have set, under Coming up, soonest first, or on Nothing set. when you have none. New reminder is the group under that, and History under that again lists the ones that have already gone off. A reminder takes words of its own, a day, a time, and a repeat: Once, Every day, Weekdays or Every week. A repeating one works its next time out from the wall clock as it fires, so it survives travel and the clocks changing.

Most reminders are about nothing in particular, which is why they are their own thing rather than a field on a task. When one IS about a task, start it from that task instead: Remind me… on the task's own menu opens the same composer already saying what the task says. Either way the task keeps its own date, which may be a different day or no day at all.

A reminder interrupts you once and is then over, so there is nothing to mark done and nothing to snooze. Arriving is what finishes it, and it moves from Coming up to History. Before it goes off, the row's menu offers Edit and Delete, which is the way to call one off. On the phone, a reminder the app has already seen arrives from an alarm the phone set for itself, which is why setting one asks for notification permission. Turning This phone and Reminders on under Notifications is what lets the server send one the phone has not seen, such as a reminder you set in a browser since you last opened the app. In a browser, Try one now under the composer asks that browser to show one, so you can see whether it reaches you there.

Recurring tasks, a row on the Tasks page under Apps, holds the tasks that arrive on their own. A row reads like a task row, carrying the status its tasks are born in, its words, and the area they are filed under, with when it comes under them — Every day, Every week, Every month, or the days of the week, read back as Mondays and Thursdays. A job writes one task per definition per day it is due, and it writes only the current day, so there is one of each on your list at a time and none stacked up on the days between.

Every fact on the row is the control that changes it. Press the pill for the status, the area word for the area, the name to rename it, and the line underneath to change when it comes: Every day, Every week and Every month are alternatives, and Certain days opens the week, where the seven days are a set you turn on and off. Pause stops a definition without deleting it and the row then reads Paused; Resume puts it back. Delete asks once and takes the definition away. Tasks it already wrote stay on your list either way — they are yours now, and nothing here reaches back for them.

+ Add at the right of the header opens a row at the top of the card with the cursor already in it. Type the name, pick when it comes, and add it. A task made this way is an ordinary task, filed under the definition's own area and type, and it reads on your Tasks list like anything else. The page is under Apps → Tasks in a browser and in the phone app alike, with the same rows and the same controls on each.

MCP, a row under Apps, opens with one line saying what the page holds: each card is an assistant connected to LifeOps over MCP, and what it wrote. Under it is every conversation that has reached your data through the connector in the last thirty days, newest first. A card holds the assistant by the name it gave when it connected, then when it was, then what it wrote under the names those rows have now. A client that gave no name reads as Connector. A conversation that only read says Wrote nothing.

Disconnect cuts an assistant off. It is a button on the connection’s own page and an item in the card’s menu, on a right-click or a hold, and it asks once before it acts. It takes back the key the assistant used to reach your LifeOps, whether the sign-in page handed it out (claude.ai and ChatGPT) or you pasted it in, so the assistant has to connect again to come back. Every other connection that used the same key is cut off with it. The card stays and says Disconnected with the time, and its list of calls stays readable, so you can still see what it did. What it wrote stays too: every task, note and entry is still yours, with the line saying which connection wrote it.

Disconnect refuses, and says why, in two cases. When the assistant used the same key this browser runs on, taking it back would sign you out as well. When the connection is older than LifeOps recording which key each one used, there is no key to take back from here.

Opening a row lists every call the connection made, in the order it made them, reads included: the time, the tool, and one line saying what it asked for. A call that wrote something has Open beside it, which goes to the row it made. The same id is on the line under a task the connector added, in the Added by your assistant band on Tasks and on the task’s own page, so a task you do not recognise leads here in one press. The list and the transcript are under Apps → MCP in a browser and in the phone app alike, with the same line under the title, the same three lines on a card, and in the phone app Disconnect on a hold of the card and on the connection’s own screen. One switch sits above the cards, My assistant acts as me, and it starts off. Off, a task a connected assistant writes waits under Added by your assistant on Tasks until you add it. On, it goes straight onto your list, still marked with who wrote it, and the assistant can add or discard what is already waiting there itself. It is for an account whose holder is the assistant. The phone app has the same switch on its MCP screen. Connecting Claude covers what a connected assistant can reach and how its writing is marked.