Omniscio

More settings (T–Z)

Reference for every Omniscio setting whose key starts with a letter from T through Z, with each section linking back to the exact camelCase key it documents.

On this page

Reference for every setting whose key starts with T through Z. Each section links back to the exact camelCase setting key.


Tasks outliner

Retired. This switch belonged to the original Tasks outliner, which was removed from Omniscio along with its tasks.md file on disk (see Tasks (personal outliner) — removed). The switch itself is gone too: Settings → Features shows no "Enable Tasks Outliner" row and Settings search has no entry for it, and a saved config that still carries the old value is ignored and never rewritten. The "needs you" boundary number that once sat beside it is retired the same way. The current Tasks feature is turned on under Settings → Features → Tasks — see Tasks. Default: off.


Tasks keyboard shortcuts

A map that lets you rebind every command-mode key in the Tasks outliner: move focus up and down; move a task up, down, left (outdent) or right (indent); fold or unfold its sub-items; cut and paste a task; move it to another project; and set its context, snooze, deadline, priority, estimate or waiting-on. The same map covers completing a task and dropping it, adding it to Today, launching or opening its agent, breaking it down, the command line, inserting a link, finding tasks, select all, and opening the cheat-sheet. Each entry maps a command name to one or more key strings. Change only the keys whose default bindings clash with your other shortcuts; the rest stay as-is. An AI helper can list, rebind or reset one command at a time with /tasks-v2/shortcuts.

The outliner's structural keys (Enter, Space, Esc, Tab and the arrow keys, with or without a modifier) can't be bound to a command: pressing one while setting a key is ignored, and one written into the map another way never takes over the outliner key. The same goes for the inbox keys Ctrl+1 and Ctrl+Q and the built-in undo, redo, copy, delete and context-menu keys: Settings and /tasks-v2/shortcuts both refuse them so a saved key is always one that works. Each person on a shared computer has their own choice.


Show all projects in one scroll

Chooses how the Tasks main panel lays out your projects. Off (the default), Tasks shows one project at a time: it opens on the view you had open last, or Inbox, and the "All Projects" entry is hidden from the sidebar and the header Show menu. On, Tasks opens on one long scroll with every project stacked under its own heading, and the "All Projects" entry comes back. Flipping it takes effect at once. It only changes the layout, never your tasks. You can find it in the Customize panel inside Tasks and in Settings under Tasks (Tasks display; search Settings for "all projects").

Default: off. This applies to everyone on this computer.


Break-it-down helper

Controls the "Break it down" helper in Tasks. When it is on, you can pick Break it down from a task's menu, press its shortcut, or choose it when you start an agent on a task, and the AI helps you split a big task into smaller steps. When you turn it off, all of those entries disappear, and Omniscio refuses the request if anything else (for example an AI helper) tries to start a breakdown. Turning it off also switches off the Looks-big nudges, because they only point to this helper. You can find it in Settings under Tasks (AI planning), and in the Customize panel inside Tasks. Tasks is still in development and is turned on under Settings, Features, Tasks.

Default: on. This applies to everyone on this computer.


Looks-big nudges

Shows a small "Break this down?" hint next to the task you have selected when that task looks too big to do in one go. A task counts as big when its text is long (about 80 characters or more), or when it is a bit shorter but is marked high importance or has a due date, and it has no open sub-tasks yet. Very short tasks never get a hint. Clicking the hint opens the Break-it-down helper. The hint appears in your project lists, not in the Today view. This setting only works while the Break-it-down helper is on. You can also switch it from the More menu in Tasks.

Default: on. This applies to everyone on this computer.


AI next-task pick

Tasks always works out a "Right now" pick for you using its own built-in ranking, which is free, instant and never leaves your computer. When this is on, a small AI request also looks at your open tasks and refines which one to do next, and it can add a short reason. Your task text, your list and project names and the names of people you are waiting on are sent first to your own Claude account (Anthropic); only when that account cannot answer do they go to OpenRouter, an outside AI service, or to whichever saved key or credits you set up under Settings → My API Keys. The request runs only when you press Re-rank on the Right now card or What should I do next? on an empty Today; the card itself never asks the AI on its own. When the AI cannot answer, Tasks falls back to the built-in pick. The AI answer is saved and reused until your tasks change, so it is not asked again on every refresh. Turn this off to use the built-in ranking only, with nothing sent anywhere.

Default: off. Turning it on is your choice, and it is stored for the account you are signed in to.


Model for arranging your day

Which Claude model does the "arrange my day" pass in Plan My Day. Leave it empty (the default, shown as Automatic) to use your default model when that is a Claude model, and otherwise Sonnet 5. Only Claude models are accepted here; anything else falls back the same way. It affects this one AI request only. When the request runs on your own Claude account (the first thing Tasks tries), it uses your default Claude model and this choice is not used. Without an OpenRouter key of your own, the fallback runs on the company lane, which may use its own approved model instead, so the model shown is not always the one that runs. Find it in Settings under Tasks (reach it by searching Settings). This applies to everyone on this computer.


Right now card

Shows a "Right now" card at the top of the Today view with the task Tasks thinks you should do next. The card has a "Not now" button to see the next choice and a Re-rank button that asks the AI to pick again. Turn this off to hide the card and stop Tasks from fetching the pick for it. Your tasks themselves are not changed. This is separate from the AI next-task pick, which only decides whether AI helps choose the task.

Default: on. This applies to everyone on this computer.


Evening wrap-up

When this is on, Tasks opens the end-of-day review for you on its own. It happens when you are on the Today view in the evening (from 6 pm by default), you still have tasks on today's plate, you have not already wrapped up the day, and you are not in the middle of typing. It opens at most once a day, and snoozing the review holds it back until the snooze ends. Turn this off if you do not want the review to appear by itself. The "Close out my day" button on the Today view still works either way.

Default: on. Each person on a shared computer has their own choice.


Evening wrap-up note

On by default. Once a day at your wrap-up hour, one note arrives in your inbox with how today went and what is coming tomorrow. It is skipped on a day you have already wrapped up, and it does not fire while Tasks itself is turned off. Turn it off to stop the note; the Today view's "Close out my day" button keeps working either way. Find it in Settings under Tasks (reach it by searching Settings). Each person on a shared computer has their own choice.


Evening wrap-up hour

The local hour, from 0 to 23, at or after which the evening wrap-up note is generated. The default is 18 (6 pm). The time zone on your computer decides which local hour this is, so it follows you if you travel. It only matters while the evening wrap-up note is on. Find it in Settings under Tasks (reach it by searching Settings). Each person on a shared computer has their own choice.


Use my calendar

Lets the Today view and Plan My Day read your Google Calendar, so meeting times can be planned around instead of planned through. When this is on and a calendar is connected, Plan My Day tells the AI when your meetings start and end; their names are sent only if "Show meeting names in AI planning" is also on, and otherwise every meeting reaches the AI as a generic "Calendar event". Turn this off to keep your calendar out of Tasks entirely — Today and Plan My Day then plan without it. It does nothing until you have connected a Google Calendar and switched the Google Calendar integration on; with that integration off, Today and Plan My Day never read your calendar or send it to the AI, even when Google is connected for Gmail. You can find it in Settings under Tasks (Tasks display), and in the Customize panel inside Tasks.

Default: on. This applies to everyone on this computer.


Show meeting names in AI planning

Decides what the AI sees about your calendar when Plan My Day builds a plan for you, and when AI arrange arranges your day. When this is off, the AI is told only that you have a meeting and when it starts and ends; every meeting is shown as a generic "Calendar event", and a meeting you typed into Tasks yourself reaches it as a generic "Task meeting". When this is on, the meeting titles are sent as well, which can help the AI plan around what each meeting is. Leave it off if your meeting titles are private. For calendar meetings this only matters when a calendar is connected and Use my calendar is on; a meeting you typed into Tasks yourself follows this switch either way, and AI arrange is told only when each meeting starts.

Default: off. This applies to everyone on this computer.


Urgency

Adds a second rating called Urgency next to Importance. When this is on, the Task details pane shows an Urgency dial beside the Importance dial, and each task row shows an urgency chip if you have set one. Importance is how much a task matters; urgency is how soon it needs attention. Turning it off hides the dial and the chip. Any urgency values you already set are kept and come back if you turn it on again.

Default: off. This applies to everyone on this computer.


Automatically share all files with agents

Controls which files attached to a task are shown to an agent you start on that task. When this is off, only the files where you ticked "Share with agent" are offered. When this is on, every file attached to the task is offered, and each file's checkbox shows as ticked and cannot be changed until you turn this off. The agent is told each file's name, type, size and location, and can open them. At most the 10 newest shared files are listed. Find it in Settings under Tasks (reach it by searching Settings).

Default: off. Each person on a shared computer has their own choice.


Only catch from these projects

Limits "Catch follow-ups when a session ends" to the projects you choose. When the list is empty, follow-ups are caught from sessions in every project. When you pick one or more projects, sessions in any other project are skipped, and a session that belongs to no project is skipped as well. Use it to keep Caught focused on the work you care about. It does nothing unless catching follow-ups is turned on. A project you chose that is later removed stops counting. If every project you chose has been removed, nothing is caught at all — the filter never widens to every project on its own, and the app logs a warning — so pick your projects again (or clear the list with "Catch from every project") after deleting them. The "Catch from every project" button clears the list. Find it in Settings under Tasks (reach it by searching Settings).

Default: no projects picked, so every project is included. Each person on a shared computer has their own choice.


Look back for missed sessions

When "Catch follow-ups when a session ends" checks for sessions it missed — because the app was closed or a scan failed — this says how many days back it looks. Between 1 and 14 days; the default is 3. Raise it to recover older follow-ups, lower it to keep the sweep tight. It does nothing unless catching follow-ups is turned on. Find it in Settings under Tasks (reach it by searching Settings). Each person on a shared computer has their own choice.


Today list name

The name shown for the built-in Today list in the Tasks sidebar and at the top of the Today view. Leave it blank to use "Today". To change it, right-click Today in the Tasks sidebar, or use its "..." button (or long-press on a phone), and type a new name of up to 120 characters. Clearing the name and confirming puts "Today" back. There is no field for it on the Settings page.

Default: blank, which shows "Today". Each person on a shared computer has their own choice.


Today list color

The color of the icon for the built-in Today list in the Tasks sidebar. To change it, right-click Today in the Tasks sidebar (or use its "..." button, or long-press on a phone) and pick a color from the palette. There is no field for it on the Settings page.

Default: no color, so Today shows its normal calendar icon in the standard color. Each person on a shared computer has their own choice.


Today list icon

The icon for the built-in Today list in the Tasks sidebar. To change it, right-click Today in the Tasks sidebar (or use its "..." button, or long-press on a phone) and pick an icon or emoji from the picker. There is no field for it on the Settings page.

Default: no icon chosen, so Today shows its normal calendar icon. Each person on a shared computer has their own choice.


Caught list name

The name shown for the built-in Caught list, where Tasks puts follow-ups it spotted at the end of your sessions. The name appears in the Tasks sidebar and at the top of the Caught view. Leave it blank to use "Caught". To change it, right-click Caught in the Tasks sidebar, or use its "..." button (or long-press on a phone), and type a new name of up to 120 characters. Clearing the name puts "Caught" back. The Caught row only appears in the sidebar when there is something to review. There is no field for it on the Settings page.

Default: blank, which shows "Caught". Each person on a shared computer has their own choice.


Caught list color

The color of the icon for the built-in Caught list in the Tasks sidebar. To change it, right-click Caught in the Tasks sidebar (or use its "..." button, or long-press on a phone) and pick a color from the palette. The Caught row only appears when there is something to review. There is no field for it on the Settings page.

Default: no color, so Caught shows its normal sparkle icon in the standard color. Each person on a shared computer has their own choice.


Caught list icon

The icon for the built-in Caught list in the Tasks sidebar. To change it, right-click Caught in the Tasks sidebar (or use its "..." button, or long-press on a phone) and pick an icon or emoji from the picker. The Caught row only appears when there is something to review. There is no field for it on the Settings page.

Default: no icon chosen, so Caught shows its normal sparkle icon. Each person on a shared computer has their own choice.


Collapse the Tasks sidebar

Remembers whether the left sidebar in Tasks (the one with All Projects, Today and your project list) is folded away. Use the "Collapse sidebar" button at the top of that sidebar to fold it into a thin strip, and click the strip ("Expand sidebar") to bring it back; Omniscio remembers your choice the next time you open Tasks. When the Tasks panel is narrow, the sidebar also folds on its own, and clicking the strip opens it again. Phones do not show this sidebar. There is no field for it on the Settings page.

Default: off, so the sidebar is open.


Team Chat DM inbox view

Retired. This option used to turn the direct-messages area of the Team Chat sidebar into a triage inbox with a People/Threads toggle. It no longer does anything: the sidebar always shows one list of people (click a person to see their Topics), and replies you owe appear in the Needs you view.


Clear Needs reply when you open the conversation

When on (the default), a direct message stops showing as needing a reply as soon as you actually look at it — the conversation is open and the App window is in the foreground. The conversation stays in your DM list either way: looking at someone never removes them from the sidebar. A new message from that person brings the tag straight back, so a glance can never bury a conversation, and a thread row clears only once you open that thread. Turn it off to go back to confirming each DM by hand with "Mark as done". Find it under Settings → Notifications → Team Chat.


Team Chat message style

Controls how Team Chat lays out messages. "bubbles" (default) places your messages on the right in rounded 3-D bubbles, like a phone messenger. "clean" uses a single calm stream with no visual distinction between senders. "compact" packs messages tighter for high-volume channels. Find it under Settings → Appearance.


Show Team Chat notices on this computer

On by default. Turn it off and new Team Chat messages stop appearing in this computer's Inbox and Needs you list, and conversations are not marked as needing a reply here; you can still open Team Chat and read and answer everything on this computer. When you turn it off, any Team Chat notices already sitting in this computer's inbox are cleared. It is meant for a computer you do not use for Team Chat, so a second machine stays quiet while your main one keeps its notices. Find it in Settings → Team Chat.


Reactions to your messages

Off by default. Turn it on and someone reacting with an emoji to one of your Team Chat messages raises a quiet row in your inbox. Never fires for your own reaction, for a channel you have muted, or while you are actively looking at that channel. It is an inbox row only — it never pops a desktop notification. This is a single global switch: you cannot turn reaction notices on for some channels and off for others.

Replies to your messages

When on (the default), a thread reply on a message you wrote — or one that @-mentions you — raises a targeted “someone replied to your message” row in your inbox. Never for your own reply, a muted channel, or a reply that does not involve you. Turning it off silences thread-reply notices everywhere: unlike top-level message notifications, this one is all-or-nothing and cannot yet be set per channel.

Quote on select in Team Chat

When on (the default), highlighting text inside a Team Chat message shows a small "Quote" button next to your selection. Clicking it drops that passage into your message box as a formatted quote with attribution, ready to reply. Off means no button appears on highlight, though the full-message "Quote" option in the message menu always stays available.

Team Chat topic preview style

Controls how the preview line under each topic is written in the Team Chat Topics list. "Inherit from session setting" (the default) simply follows whatever you picked for session titles — nothing to keep in sync. The explicit choices are Brief (first line only, up to 60 characters), Descriptive (the full preview text as written), plus Journey, Latest Topic, and Custom, which currently behave like Descriptive and are kept for upcoming author-trail and custom-formatting options. Find it under Settings → Notifications → Team Chat.


Team Time

A world-clock dashboard for distributed teams. Shows live clocks for each teammate's timezone, their current working-hours status (in/out of business hours), and a meeting-planner that highlights times when everyone overlaps. Turn it on to add the Team Time panel to the sidebar. Default: off.


Telegram engine

Chooses which Telegram client library Omniscio uses behind the scenes: "mtcute" (the default — the actively maintained MTProto engine) or "gramjs" (the older engine, available as a fallback). Switching engines signs you out of Telegram: each engine keeps its own session, so you reconnect once, and the app needs a restart before the change takes effect. Your chats and messages are not deleted, and switching back leaves the other engine's session untouched. Change it only if you notice connection issues with the default. Found in Settings → Channels → Telegram engine.


Text correction model

The AI model used for grammar and spelling fixes — the Hotstring Creator suggestion and Instant Fix. Leave it empty or set "Automatic" (the default) for a fast built-in default. Set a specific model ID to override for this feature only. Found under Settings → Sessions → Per-feature AI models.


Keep typing in the box you are typing in

Once you start typing in any text box, only you can move the cursor out of it — by clicking somewhere else, or pressing Tab, Escape or Enter. If the app rebuilds the box or switches it off for a moment while you type, your next letters still go into the box (or are dropped), and never trigger a one-letter shortcut such as R, N, I or J. Turn it off to go back to the old behaviour. Find it in Settings > Keyboard. Default: on.


Tool update notifications

When on, Omniscio shows the toast pop-ups that announce "tool update available" for connected tools like Git. Background version-checking still runs either way, and updates remain visible and installable from Settings → Connected Tools — this setting only controls the notification. Default: off, so you are not interrupted unless you ask to be.


Trusted folders

A list of Google Drive folders you have authorized for two-way read/write access. Each entry records the folder ID, a label, whether its subfolders count, and which columns are editable. A spreadsheet inside a trusted folder needs no entry of its own. The folder is checked with Google Drive every time an assistant writes, so moving a sheet out of the folder takes its access away at once, and a check that cannot get an answer refuses the change. Checks need a Drive permission Google treats as restricted, so this app only asks for it while the in-development Lab feature Trusted Sheets folder checks is switched on; while that is off, folder entries grant nothing and a write into one is refused with a message saying so. The Use trusted folders switch beside this list turns every folder entry off without deleting them; entries on the Trusted Sheets list keep working.


Trusted sheets

A list of Google Sheets you have authorized for two-way read/write access. Each entry records the spreadsheet ID, a human label, which columns are editable, and an optional task-view mapping (which columns hold subjects, due dates, completion status, and so on). Add a sheet here before Omniscio can write back to it; read-only access does not require an entry.


Trusted Docs

The list of Google Docs an assistant is allowed to change. An assistant can change a Google Doc only when it is on this list; reading is not affected, so it can still read every document it can reach. Paste a Doc link or its document ID, optionally give it a name, and add it — only you can add documents here. The list holds up to 100 documents, and removing one takes its write access away at once. With no documents on the list, assistants cannot change any of your Google Docs. Find it in Settings → Google Workspace.


Voice confirmation speed

Playback speed for the short spoken confirmations Omniscio reads after you issue a voice command ("Got it", "Done", and similar). Accepts a multiplier between 0.5 (half speed) and 2 (double speed); default 1.2. Increase it if the confirmations feel slow; decrease it if they are hard to follow.


TTS summarization threshold

The character count that decides whether a text-to-speech response is read in full or summarized first. Responses shorter than this number are spoken verbatim; longer ones are condensed before being read aloud. Accepts a value from 100 to 5000 characters; default is 500. Raise it to hear more responses in full; lower it to keep audio brief.


Typing speed insights

Passively measures your real typing speed in the session message composer and surfaces it as a "Typing speed" stat in the Statistics panel. No AI is involved and the data never leaves your machine. Default: on. Turning it off stops all sampling immediately.


Interface language

Sets the display language for the entire Omniscio interface. Choose from the supported locales — English (default), Spanish, Japanese, German, French, Arabic, and many more. Pilot languages (Tagalog, Traditional Chinese, Turkish, Urdu, and others) are also listed when their pilot flag is on. Any screen that has not yet been translated falls back to English automatically.


Update everything

Turns the "Update everything" toolbar action on or off. When on, one button checks and updates your Omniscio app, your command-line tools, your skills, and your plugins together, instead of you hunting through each one. Default: on. Turn it off to hide the action and update each piece on its own.


Update everything automatically

When on, the whole-install update runs by itself on launch and every six hours, and stays quiet unless something needs you. It is the same update the "Update everything" toolbar action runs, just unattended. Default: on. It has no effect while "Update everything" is off, since there is then no action to run.


Usage cascade

A fallback list for when your AI usage runs out. When every Claude account you have is at its usage limit, or a session on another provider hits its limit or runs out of balance, that session moves to the first fallback on your list that is switched on, set up, and not already full. The session's last turn then runs again there, and a note in the session says it moved.

For each fallback you can pick the model it runs, cap how many sessions may use it at once, switch it off without removing it, and change its order. Claude itself cannot be a fallback. Some fallbacks bill per token, and the list marks those. The list starts empty, and Omniscio never adds anything to it for you. With nothing on it, sessions simply wait for the limit to reset. Find it in Settings > Accounts. Default: empty.


Usage forecast

Shows a real-time prediction of how much time you have left before your Claude usage-limit window is exhausted, based on your recent consumption rate. The forecast appears in Settings → Notifications and as a compact indicator elsewhere in the app. Turn it off if you do not need the estimate. Default: on.


5-hour exhaustion warning

How many hours before your 5-hour Claude usage bucket is predicted to run out that Omniscio should warn you. Set to 0 to disable this specific warning. Maximum value is 5 (warn up to 5 hours ahead). Only meaningful when the Usage Forecast is on. Found under Settings → Notifications.


Weekly exhaustion warning

How many hours before your weekly (7-day) Claude usage limit is predicted to run out that Omniscio should warn you. Set to 0 to disable this warning. Maximum value is 168 (the full week). Only meaningful when the Usage Forecast is on. Found under Settings → Notifications.


Usage widget — show the 5-hour window

Shows your 5-hour usage window — how much of it you have used and how long until it resets — inside the optional Usage widget in the top header bar. On by default. It only does anything once you have actually added the Usage widget to the header, from the toolbar overflow menu or Settings → Widgets. Turn it off to drop just the 5-hour line and keep the rest of the widget.


Usage widget — show reset countdowns

Shows how long is left until each usage window resets, next to its percentage, in the optional Usage header widget. On by default; turn it off to show just the percentages and keep the widget narrower. Only relevant once the Usage widget is in your header.


Usage widget — show the weekly window

Shows your weekly (7-day) usage window — how much of it you have used and how long until it resets — inside the optional Usage widget in the top header bar. On by default. Like its 5-hour sibling, it only does anything once you have added the Usage widget to the header. Turn it off to drop just the weekly line.


What you do

A short note about what you do, up to 80 characters, for example "founder" or "designer". Omniscio asks for it once during setup, and you can change or clear it here any time. It is optional and stays on this computer. Right now it is only a note about you. Nothing in the app changes its behavior based on what you type, so leaving it empty costs you nothing. Find it in Settings > Diagnostics (Health tab), under "What do you do?". Default: empty.

Vendor key watch alert threshold

How many dollars of unexplained spend a vendor must show before the vendor key watch raises a card. Default 5. The gap must also be more than a fifth of what Omniscio recorded, so rounding between price lists never alarms. Set 0 to turn the cards off while the Stats → Spend rows stay live. Only relevant when vendorKeyWatchEnabled is on. Number; accepted values: 0–10000.

Vendor key watch shared keys

The vendors whose API key other computers also use. A marked vendor is never compared and never raises an unexplained-spend card, because this computer cannot see the other computers' spend; its Stats row reads "Not watched" with that reason. Unmark it and the vendor returns to the watch on the next hourly check. Other vendors are unaffected. Default []. List of vendor ids, at most 32; for example ["deepseek"].


Visual theme

Picks the decorative skin applied on top of the base dark or light mode. Built-in choices include Glassmorphism (frosted glass panels), Linear (clean flat borders), Aurora (soft gradient accents), Neon, Spotify-green, Luxury, and several others. You can also import a custom .theme.json file. Changing this setting takes effect immediately without a restart.


Enable Video Studio

Adds Video Studio to your sidebar, where you make short AI video clips and fix any part of them. Clips are paid for with Video Studio credits, and you connect your Video Studio account from the panel itself. Desktop only. It also needs the Video Studio switch under Settings → Lab turned on; until that early-access switch is on, the feature is hidden and this one has nothing to show. Off by default. Find it in Settings → Built-in Apps.


Command capture mode

Controls what the microphone does after you speak a voice command. There are four choices:

  • Auto uses Single command when a session is open and Continuous when none is.
  • Single command stops the microphone after each command.
  • Continuous runs each command as soon as you say it and keeps the microphone open for the next one.
  • Buffered keeps what you say and waits until you press Execute before running it.

This only applies to voice commands. Dictation, where your words are typed for you, ignores it. Find it in Settings > Voice, once voice is turned on. Default: Auto.


Ask about a thread (voice)

Lets you ask Omniscio about a specific session by voice — hear a spoken summary or ask a question about what an agent did, hands-free. This is voice access to session context, not a live conversation with the running agent. Requires a voice key to be configured. Default: off.


Ask the live agent (voice)

Lets you queue a spoken question to a session that is actively running and get notified when the agent answers. Your question is routed into the session and the reply is read back to you when it arrives. Requires a voice key. Default: off.


Voice recognition language

The language the speech recognizer listens for, picked from a short list of plain language codes — "en" (English), "es", "fr", "de", "ja", "ko", "zh", "pt", "it" and the rest of the list in Settings. The default is "en", not your system locale, so someone whose computer is set to another language still has to pick theirs here the first time. A short string up to 10 characters.


Voice report back

When on, Omniscio reads the AI's response aloud after each voice command completes, so you can hear the result without looking at the screen. Works with whichever TTS provider you have configured. Default: off (opt-in). Turn it on for a fully hands-free loop.


Voice report back character limit

The maximum number of characters of an AI response that Voice Report Back will read aloud. Responses longer than this are truncated. Accepts a value from 50 to 10 000 characters. Lower it to keep audio brief; raise it when you want the full answer spoken. Only has an effect when Voice Report Back is enabled.


Auto-stop silence timeout

How long the microphone waits in silence before it stops listening on its own. Omniscio watches the volume from your microphone, and when it stays quiet for this long, listening ends. You can set it from 0 to 10 seconds, in half-second steps. Set it to 0 if you want to stop the microphone yourself every time.

The slider is hidden when Command capture mode is Single command, because the microphone already stops after each command in that mode. Find it in Settings > Voice. Default: 3 seconds.


Wait confirmation

When on, a session turn that ends with ambiguous waiting language (phrases that suggest the agent is waiting but are not clear-cut) triggers one hidden confirmation probe: Omniscio asks the agent to clarify before deciding whether to flip the session to "Needs You". If the agent confirms it is waiting, an automatic keep-alive is issued; otherwise the original message surfaces normally. Each probe is bounded to one round and the exchange is invisible. On by default; turn it off if you would rather not pay an extra agent turn for each ambiguous wait.


Wake word engine

Which local wake-word detection engine listens for the trigger phrase. "openwakeword", the open-source engine, is the only one you can pick, and it runs entirely on-device with no audio sent to the cloud. A second engine, "picovoice", is retired: its models are no longer bundled, it is hidden from the picker, and a saved "picovoice" choice quietly runs openwakeword instead. There is nothing to switch to today; if the trigger phrase misfires, adjust the wake word itself rather than this field.


"You don't need to watch" nudge

When on (default, desktop only), Omniscio shows a small dismissible hint inside a session if it detects you have been watching a running agent for roughly 45 seconds — window focused, mouse still moving — reminding you that the session will come to your inbox the moment it needs you. Appears at most once per session. Turn it off under Settings → Notifications if you find it distracting.


Watch-time recap card

When on (default), Omniscio adds a positive inbox card — at most once a week and only after a day you spent a lot of active time watching agents work — showing roughly how much time you spent watching and encouraging you to run more sessions in parallel. The analysis runs entirely on your machine with no AI. The card includes a "Turn off these notices" button. Turn it off via Settings → Notifications.


Mobile server bind interface

Determines which network interface the Omniscio mobile-access server listens on. "127.0.0.1" restricts it to this machine only (safe default for tunnel-only setups). "0.0.0.0" exposes it on all interfaces, including your local Wi-Fi, so you can reach it from a phone on the same network without a tunnel. Only change this if you know your network is secured. Found under Settings → Web Access → Advanced.

The radio buttons are disabled while mobile access is running, and the control that stops it is the Mobile Access toggle higher up the same page — there is no button named "Stop". You rarely need to come here at all: when the server is listening on this machine only and no tunnel is carrying it, the Scan to Connect popover says so and offers a one-click Let my phone connect over Wi-Fi that writes this setting and restarts mobile access for you.

Leaving this on "127.0.0.1" does NOT break Tailscale. tailscale serve proxies to http://127.0.0.1:<port>, so a tunnel reaches the server regardless; the setting only governs plain same-Wi-Fi pairing, where the phone dials the computer's LAN address directly. Widening it is also not always sufficient on its own — Windows still has to allow inbound connections on the port, and a Wi-Fi network classified Public blocks them harder than a Private one.


Weekly summary cost cap

A spending limit in US dollars that caps how much the Weekly Summary feature can spend on AI calls on any single day. Accepts a value from $0.05 to $5.00; default $0.10. If the cap is hit, the summary for that day is skipped. Adjust it to stay within your overall AI budget.


Include Omni

On by default. When Omni is on, the Weekly Summary adds one line saying how much Omni did that week: form answers sorted, document action items filed, spreadsheet reviews and issues flagged. The line is hidden when Omni did nothing, and only the counts reach the summary's AI. With Omni off this switch does nothing.


Weekly summary scheduled hour

The hour of day (in your local time, 0–23) when the Monday weekly summary generates. Default is 7 (7 AM). Set it to whichever hour you are most likely to read the summary. The minute can be configured separately via the weekly summary scheduled minute setting.


Enable Where's My Work

Off by default. Turn it on to add a Where's My Work icon to the toolbar. It opens a read-only screen showing what actually became of your changes — which branches landed, which are waiting in review, which were declined, and which are stranded with nobody carrying them forward. It only reads; it never moves or merges anything.

Auto-continue self-identification

When on, Omniscio's automatic "keep going" nudge — the bare "Please continue." message it injects when a session is auto-waiting — prefixes itself as a system message so the agent knows it came from Omniscio, not from you. This prevents agents from treating a terse user message as a background wake signal and ignoring it. Default: off (opt-in). The manual Nudge button is never prefixed.


Word export

When on, an "Export to Word" option appears in session and document export menus alongside Markdown and PDF. Default: off. Enabling it does not send anything to an external service — the conversion happens locally. Turn it on if you regularly need .docx output from sessions or the Writer Studio.


Writer inline autocomplete

Turns on ghost-text inline suggestions in the Writer Studio editor — as you type, a faded continuation appears that you can accept with Tab. Default: off, because autocomplete fires on every keystroke and can add noticeable AI cost if you use a premium model. Enable it once you have set a cost-effective autocomplete model.


Writer autocomplete model

The AI model used only for the Writer's inline autocomplete (ghost text), separate from the model used for edits and the Assistant chat. Defaults to the cheapest available model (Luna) when left empty. Set a different model only if you want higher-quality completions and are comfortable with the extra cost, since autocomplete runs constantly as you type.


Writer daily cost cap

Retired. This setting used to cap what Writer Studio's AI could spend in a day. It no longer does anything: Writer's AI runs only on your own connected AI (your Claude sign-in first, then your own API keys in your My API Keys order), never on Omniscio credits, so there is no Writer daily budget and no "Writer AI edits" row in the Cost Control panel. A saved config that still carries the old value is ignored and never rewritten. What Writer's AI costs is still recorded for the spend readout. The one-time Writer credits notice setting (writerCreditsNoticeAccepted) is retired the same way.


Writer AI model

The AI model used for the Writer Studio's Assistant chat, selection edits, and rewrite suggestions. Leave it empty or "Automatic" to use the built-in defaults (Sonnet for chat, Luna for edits). Set a specific model ID to override all three at once. Autocomplete has its own separate model setting. Found under Settings → Writer.


Writer page palette

A curated color palette applied to the Writer page background and editor surface, independent of the app's visual theme. Options include "none" (plain), "parchment", "midnight-ink", "candlelight", "noir", "velvet", "frost", "sage", and "rose". Default is "none". A purely visual preference — pick whichever helps you focus.


Writer sidebar

Shows or hides the Writer Studio entry in the main sidebar. Turning it off removes the Writer row from navigation without deleting any documents. Turn it back on at any time to restore it. Default: on.


Writer editor theme

Controls the color scheme of the Writer Studio editor itself, independently of the app-wide dark/light setting. "match-app" (default) follows whatever the app is using. "light" forces a white editor. "dark" forces a dark editor. Useful when you prefer writing in a light editor while the rest of the app stays dark.


Writer toolbar layout

Controls how the Writer Studio document header toolbar buttons are displayed. "Icons only" (default) shows just icons with rich hover tooltips. "Icons + labels" shows icons alongside full text labels. "Compact" shows icons with shortened labels. "All icons" makes every toolbar button icon-only, including File, Share, and Send-to menus.


Freeze-proof mobile connection

Keeps your phone's connection to Omniscio alive even if the desktop app freezes briefly, by answering your device's WebSocket from a separate worker thread. Falls back automatically if it cannot start. Takes effect after an app restart. Default: on. Turn it off only if you suspect the worker transport is causing trouble — the app then serves your phone from the main thread, the way it did before.


Notices from the tray

Sends Omniscio's own notices — an outage, a scheduled job failing, a budget or balance problem — from the tray icon, so they read apart from your session alerts. Windows only: the tray balloon it uses is a Windows API, so on macOS and Linux the notices still arrive as ordinary pop-ups and only their label separates them from your session alerts. Off by default; your session alerts are never affected either way. Read when a notice fires, so flipping it needs no restart.


Custom voices

The voices you added yourself, each stored as a name and a provider voice ID. Add one by pasting a voice ID — it then appears in the voice list alongside the presets and can be chosen as your TTS voice. Removing a custom voice that is currently selected clears the selection.


TTS Voice

The voice model used for text-to-speech playback. Pick one of the presets, or use a custom voice you have added, and Omniscio speaks with it wherever it reads text aloud. Leave blank to use the provider's default voice.


Show X/Twitter posts as screenshots

When on (the default), a link to an X/Twitter post is rendered inline as the post's screenshot wherever it appears — a session message, an inbox alert, a drip item, or a team-chat message. The picture is fetched once and reused on later views. Rendering uses a paid relay call, so turn this off to keep just the plain link.


Widget Sidebar

A side panel beside your sessions holding live widgets you choose: what is waiting on you, what is running, today's spend, a note, the clock, widgets your plugins and scripts provide, and web pages you add by address. Off until you turn it on — nothing appears and nothing runs before that, and turning it off again puts everything back as it was. Show or hide the panel with Alt+Ctrl+Shift+D. Desktop only. See Widget Sidebar.