Omniscio

More settings (R)

Plain-language reference for every Omniscio setting whose key starts with the letter R, in the alphabetical settings reference.

On this page

Plain-language reference for every Omniscio setting whose key starts with the letter R.


Recipe approval timeout

When a recipe step contains an approval gate and that step does not specify its own timeout, this value determines how many days the gate waits before auto-resolving. Once the countdown expires the step is automatically closed so the recipe does not wait forever for human input. Default is 7 days; you can set any whole number between 1 and 90. Change this if your workflows routinely need more (or less) time for a human to act on an approval prompt.


Record-break inbox alerts

When turned on, Omniscio adds a card to your inbox every time you set a new milestone record in your Stats (for example, most sessions in a day or highest daily spend). Off by default — turn it on if you want a lasting inbox record of personal bests. This is the only way a record break reaches you outside the Stats tab: there is no toast for it, and the card arrives at most once a day.


Recurring special events

Lets you register birthdays, anniversaries, and holidays that automatically create Google Calendar events with lead-time reminders (for example, a day before, a week before, or a month before). Off by default and currently in development; visible only when enabled through the Labs toggle. Desktop-only for the initial release.


Reduce terminal popups

On Windows, agent commands can cause a large Windows Terminal window to steal focus momentarily. Turning this on switches the Windows default terminal to the classic Console Host so any stray window appears as a brief, unobtrusive flash instead. The change is fully reversible — turning it off restores your previous Windows terminal setting. You can still open Windows Terminal yourself at any time. Off by default; Windows-only (no effect on macOS or Linux).

Model for re-titling a session

Which AI model rewrites a session's title when the conversation has moved on. Leave it on Automatic and Omniscio picks a sensible default — a fast, low-cost model, since a refresh has the earlier title to work from and re-titling happens far more often than the first title. Pick a specific model to override that. Only used when session naming generates AI titles. Runs on the app's shared provider keys, so it does not need a key of your own.

Repeated-phrase quick-reply nudge

When on, Omniscio quietly watches for phrases you type repeatedly to your agents. Once a phrase crosses the threshold (see below), an inbox card suggests saving it as a one-tap quick reply so you never have to type it again. A phrase is only ever nudged once. On by default. Turn off if you prefer no coaching on your typing habits.


Repeated-phrase nudge threshold

The number of times you must type the same phrase before the quick-reply nudge fires. Default is 10; range is 2–100. Lower this if you want suggestions sooner; raise it to reduce nudge frequency. Has no effect when repeatedPhraseQuickReplyNudgeEnabled is off.


Reply suggestion model

The AI model used to generate suggested replies in sessions. When left on Automatic (the default, stored as an empty or null value), Omniscio inherits the global Utility AI model — the same model that powers your AI suggestions globally. Set a specific model string here to override it just for reply suggestions. Found in Settings → Sessions → Per-feature AI models.


Repo Foundations panel

Controls the visibility of the Repo Foundations dashboard card. This setting was retained for back-compatibility after the feature shipped; the card is now always visible through the normal unreleased-features gate (status: shipped) regardless of this value. Default false; the field is effectively inert and no Settings UI toggle is exposed.


Repo Foundations skill

Installs (or removes) the bundled /repo-foundations Claude Code skill. When on, you can point the skill at any repository and it inspects the tech stack, then — after you review and approve a full preview — writes a tailored documentation system, Claude-rules baseline, and a .claude/foundations.json manifest into the working tree for you to commit. Turning it off removes the skill from your Claude Code skills list. On by default (shipped August 2026).


RepoGuard builtin migrated

Internal one-time migration flag. Set to true automatically after the built-in RepoGuard feature is replaced by the marketplace plugin version for existing users. There is no Settings UI for this field; it should not be changed manually.


Approve AI Coaching profile writes

When on, an agent that wants to change your AI Coaching profile has to wait for you to approve it first. That covers creating, changing or deleting a coaching artifact, resetting an interview, deleting a goal, clearing the saved memories, and updating or regenerating your core profile. These matter because coaching artifacts become facts that every session is told about you, and several of these actions delete things. Turn it off only if you are comfortable with agents editing your coaching profile without a check. Find it in Settings > CLI Control, under Approval requirements. Default: on.


Approve archiving a memory

When on, an agent that wants to archive, restore or edit one of your saved memories has to wait for you to approve it first. Off by default, because all three of those are recoverable — an archived memory can be brought back. Turn it on if you would rather review every change to your memory store. Permanently deleting a memory always needs approval and cannot be switched off here.


Approve setting up AutoHotkey

When on, an agent that wants to set up AutoHotkey for you has to wait for you to approve it first. That one setup does several things in a row: it turns the AutoHotkey plugin on, installs AutoHotkey itself if it is missing, edits your own AutoHotkey script so it includes a generated file, and starts it. The approval card names each of those steps.

It is off by default, because the setup is meant to finish in one go with nothing left for you to do. Turn this on if you want a card before the whole setup runs. Find it in Settings > CLI Control, under Approval requirements. Default: off.


Approve handing a session off

When on, an agent that wants to hand its work off to a fresh session has to wait for you to approve it. A handoff summarizes the session — a paid AI call — and then starts a fresh successor session that carries on the work, so it both costs money and creates a new running session. Off by default: a handoff continues work you already authorized, and gating it used to stall a build halfway through while nobody was at the keyboard. Turn it on if you want to see every handoff before it happens.


Approve sending a Slack message

When on — the default — an agent that wants to post to Slack has to wait for you to approve the message first. Unlike most CLI actions this one is not confined to your machine: sending to Slack puts words in front of real people in a real channel and cannot be taken back, which is why this ships on. Turn it off only if you are comfortable with agents posting on your behalf without a check.


Approve Supermail filter changes

When on, an agent that wants to add or change a Supermail filter rule has to wait for you to approve it. A filter change is not a one-off — it changes how your incoming mail is sorted from then on. Off by default, so filter writes apply immediately — that keeps things working exactly as they did before this approval existed. Turn it on if you want to review changes to how your mail gets sorted. Sending mail and other Supermail controls have their own separate approvals, both on by default.


Require global auth

The enforce-sign-in flag for the Global Auth gate. When on (the default), the app requires you to sign in to your Omniscio account before use. This is set by the app's auth-gate provisioning and is not user-editable from the Settings UI or the CLI. Synonyms: enforce sign-in, sign-in gate, hosted auth required.


Require mobile account login

An older security gate that asked a paired phone to prove it was signed into the same cloud account before performing any powerful (mutating or spending) action. It is retired and does nothing on a normal install, whatever this setting says. Require device approval for mobile access replaced it and is checked first, so this gate stands down wherever that one is on — which is every normal install. It also could not be left on: the phone's Google sign-in looped forever over the remote-access domain and locked the owner out. It still runs as a fallback in exactly one place — a test setup where device approval has been switched off by an environment variable — and it fails closed there. The protection your phone actually has is device approval, not this. Kept for back-compatibility; switching it on does not bring it back.


Require device approval for mobile access

When a new phone connects with your web-access pairing token, it is blocked from doing anything — reads, actions, and the live feed — until you explicitly approve it, either from your signed-in desktop or with a one-time code sent to your account email. A leaked pairing token is therefore useless to anyone who does not also have access to your desktop or your email. Default: on.

The signed-in owner can switch it off in Settings → Remote Access, for the times when being away from the computer would otherwise lock them out of their own phone. Turning it off asks for confirmation first, because while it is off anyone holding the access link can open the app and read your data. It cannot be changed from the command line or from a paired phone — only by your own click on the desktop. An older saved off value is reset to on once when this version starts, since the setting was ignored for a while and a stale value must not quietly remove the protection.


Require isolation for every session

When on, a session whose own copy of the project cannot be created is stopped with an error instead of silently falling back to run inside your main project folder — so an agent can never edit or commit to your shared project tree when isolation fails (for example, under heavy load). Off by default; turn it on if you rely on isolation as a safety net. It pairs with Default isolation folder: once that is set, a session that cannot get its own copy runs from a safe scratch folder instead of stopping. Env kill switch: AMC_DISABLE_REQUIRE_ISOLATION=1 — honoured in a developer's own from-source run only; an installed build ignores it and follows this setting.


Work on separate copies

Whether a new session starts isolated — running against its own copy of the project on its own branch instead of the shared project folder, so two sessions running at once cannot overwrite each other's edits. Off by default. This is the app-wide default only: a project can carry its own (Edit Project → More options → Isolate new sessions), and a single session can override it before it starts — Shift+click the + button in the Sessions panel, or open a not-yet-started session's ⋯ menu and toggle Worktree Session. Changing it affects the sessions you start next, never the ones already running. Session isolation covers where that copy is created, what is inside it, and how the work comes back to your project.


Default isolation folder

A safety net for the one case where the copy cannot be set up: an absolute folder Omniscio may use when creating a session's isolated copy fails. It is only consulted while Require isolation for every session is on. Instead of stopping the session with an error, Omniscio copies your source into a subfolder of this path and runs the session there, so the agent still never touches your main project folder. Leave it empty (the default) for no fallback — the session then simply stops with an error, which is the fail-closed behavior Require isolation for every session exists to give you. Point it at a folder on a drive with room that your account can write to: a path Omniscio cannot create is the same as having no fallback at all.

Omniscio owns the copies it creates in this folder and prunes them: each fallback copy is recorded when it is made, and a later fallback session deletes the recorded copies older than 24 hours, keeping the newest three. Anything left in a fallback copy — uncommitted edits included — is deleted with it, so finish and move work out before then. Subfolders Omniscio did not create are never recorded and never deleted, so a folder you have used for other things is safe; for tidiness, point this at a dedicated folder rather than one you also use for other projects.


Reserve a UI core

On machines with Intel 12th-gen or newer hybrid CPUs (Performance + Efficiency cores), turning this on fences every Claude CLI session off one full Performance core, reserving it exclusively for Omniscio's own window and main process. This prevents the "window frozen under agent swarm" stall that occurs when 200+ agent sessions starve the UI of CPU time. Omniscio's own processes are not pinned to that core, so a heavy redraw can still use all cores when the box is idle. Suspended while a global CPU-Burst window is open. Windows-only; silent no-op on non-hybrid CPUs and on macOS/Linux. Off by default (opt-in).


Romanian language pilot

Adds Romanian (Română) to the Omniscio language picker as a pilot catalog translated by AI. Any string not yet translated falls back to English. Off by default; enable through the language settings if you want to try the Romanian interface.


Route internal AI through gateway

When on (the default), Omniscio's built-in AI helpers — session title generation, reply suggestions, smart categorization, the daily digest — are routed through the company's cloud gateway using your signed-in account token instead of hardcoded provider keys. On by default since July 2026. Turning it off sends those calls directly to the provider endpoint, byte-identically to the pre-gateway path. If you have set your own OpenRouter key, that key always wins and bypasses the gateway entirely regardless of this toggle.


RSS feed digest

When on, Omniscio generates an AI-curated morning briefing that summarises the new articles across all your RSS subscriptions. Two AI passes run: one to curate the most relevant items, one to produce plain-language summaries. Off by default. Requires the RSS integration to be enabled. You can control the delivery time with rssDigestTime and personalise the curation with rssDigestInterests.


RSS digest interests

A free-text field (up to 1,000 characters) where you list topics you care about — for example, "AI research, electric vehicles, product design". The digest curation pass uses this to weight which articles to surface for you. Leave it empty for unweighted curation. Only active when rssDigestEnabled is on.


RSS digest delivery time

The time of day the RSS feed digest is generated and delivered to your inbox. Stored as a 24-hour HH:MM string; default is 07:00. Change this to match the time you typically start your day. Only active when rssDigestEnabled is on.


RSS reader font size

Controls the text size used when reading articles in the RSS reader panel. Options are Small (sm), Medium (md, the default), Large (lg), and Extra Large (xl). You can also change this on the fly from the appearance popover inside the reader without visiting Settings.


RSS reader width

Controls how wide the article content column is when reading RSS articles. Options are narrow, medium, wide (the default), and full. Choose a narrower width for a comfortable reading line length; choose full to use all available space. You can also change this from the appearance popover inside the reader.


RSS article retention

How many days Omniscio keeps RSS articles before automatically purging them. Default is 90 days; range is 7 to 730. Shortening this value triggers an immediate cleanup of articles older than the new limit. Raise it if you like to catch up on a long backlog; lower it to save disk space.


Running Apps

When on, Omniscio shows a Running Apps entry in the sidebar — a live list of the dev servers it has launched through the bundled portless proxy. Each server gets a stable .localhost URL so agents and your browser can reach it consistently. Requires the portless proxy and Node 24. Off by default (opt-in); install it from Marketplace → Inbuilt if you want to see and manage your running dev servers from within Omniscio.


Russian language pilot

Adds Russian (Русский) to the Omniscio language picker as a pilot catalog translated by AI. Any string not yet translated falls back to English. Off by default; enable through the language settings if you want to try the Russian interface.


Reply to a Help Desk customer

Sending a reply from the Help Desk console over the CLI emails a real message to the customer who opened the ticket. On by default, so the reply is held for your approval before it goes out. Turn it off to let replies send without a prompt.


Reason API key

The API key used to run Reason sessions. Paste it into the Reason card under Settings > Accounts; it is stored encrypted and never shown back in full. Leave it blank to keep Reason unavailable as a session provider.


Resource Auditor can start fix sessions on its own

When the hourly Resource Auditor finds something heavy it cannot name, it may start a session that works out why and fixes it — at most one every four hours. That session costs money, so this is off until you say yes; it is on by default on a developer's own source copy of the app, which is where the hourly audit runs today. Turning it on asks for confirmation first.