Chief of Staff — the guided assistant for non-technical users
Chief of Staff turns Omniscio into a single friendly chatbox. Instead of learning projects, sessions, panels, and settings, you turn on one toggle — or just click the Chief of Staff item in the sidebar — and talk to it in plain language — "remind me to call the vet at 4", "save this as a note", "what needs my attention right now?" — and it does the work for you behind the scenes. It's built for people who want the power of Omniscio without the cockpit.
On this page
What it is
Chief of Staff turns Omniscio into a single friendly chatbox. Instead of learning projects, sessions, panels, and settings, you turn on one toggle — or just click the Chief of Staff item in the sidebar — and talk to it in plain language — "remind me to call the vet at 4", "save this as a note", "what needs my attention right now?" — and it does the work for you behind the scenes. It's built for people who want the power of Omniscio without the cockpit.
Your Chief of Staff is a real Omniscio AI assistant, not a scripted bot. Under the hood it is an ordinary Claude session — the same kind that powers Ask Omniscio — cloned and pointed at a friendlier, plain-language persona, with extra guard rails turned on because it's aimed at people who won't read the fine print.
Chief of Staff is an opt-in Lab: off by default, and announced once so you know it is
there to try. It stays hidden until you reveal it — Omniscio says so in your inbox when it
becomes available — and you turn it on from Settings → Lab (listed as Chief of Staff
Mode). The AMC_SHOW_SIMPLE_MODE=1 development flag only makes the switch reachable; it
never turns the mode on. It is still an early version, so expect the odd rough edge.
Naming: "Chief of Staff" is the user-facing name. Internally the feature is still "Simple Mode" — the
simpleModeEnabledsetting, the__concierge__session, and theSimpleMode*files keep that name (same split as KMS ⇄ "The Vault"). Only the labels the user sees say "Chief of Staff".
Where to find it
Click the Chief of Staff item in the sidebar, or turn the mode on from Settings → Lab, where it is listed under its former name.
How it behaves
- Plain language, no jargon. The concierge never shows you code, file paths, IDs, or tool names. It gives a short friendly acknowledgement, a note if something takes a moment, and a plain-English result — not a play-by-play.
- Starting a chat is instant. The moment you send your first message, your chat opens and your message shows right away, with a "Thinking…" note while your assistant spins up in the background — no blank waiting. Your words stay on screen the whole time, even while it decides where the request belongs or asks you to approve starting work in one of your projects. If the chat can't be started at all, your message goes back into the box so you can send it again without retyping. If the chat opens but never gets going, after about half a minute the screen says "This chat couldn't start" with a Try again button that starts a fresh chat with the same words — never a stuck screen.
- You can see what it's doing, in plain words. While it works the screen always moves: a plain "Working on it…" line at first, and then, as soon as it starts doing things, a short checklist — one honest line per thing it actually does, like "Checking your email" — ticked off as each finishes. The lines are read from what the assistant is really doing as it works, never guessed at (a to-do list the assistant writes in its own reply is not used), and the raw details behind them are never shown. A long job keeps the latest few lines in view and counts the rest rather than scrolling you away from them.
- If it has to wait, it tells you. When your AI allowance is used up for now, the chat shows a quiet "Taking a break" note saying it will pick up on its own exactly where it left off — so a pause never looks like a screen that has given up.
- If something goes wrong, it says so — in plain words, with one thing to do. A chat never goes quiet. When your Chief of Staff can't answer, a short card appears instead: no AI account is signed in (a Sign in button takes you to Accounts in the full app), your AI allowance is used up (a See my plan button), the chat couldn't start, it was interrupted, or it is taking longer than it should — each of those last three with Try again, which sends your last message again (as a fresh chat when the first one never started). Tapping it twice does not send twice, and your message shows once in the chat, not once per try. When Omniscio is already retrying on its own, the card just says "I'm trying again on my own" and has no button, so nothing is sent twice. Off the desktop app, Sign in and See my plan say "Desktop only", because signing in happens on your computer. If a follow-up message can't be sent, a short note appears and your words go back into the box.
- Housekeeping stays out of the chat. The automatic nudges Omniscio sends the assistant behind the scenes — like asking it for a final answer it forgot to give — never show up as messages you didn't write.
- One question at a time. When it needs a decision it asks a single simple yes/no question, never a wall of options.
- Clickable choices and a short version, like any chat. When your assistant offers you a choice, the options appear as buttons you can tap instead of typing a letter. Each reply also carries a Plain Speak card, with the same Full / Short toggle as the rest of Omniscio. Chats you started before this change keep their old look, with no card, but they learn the clickable choices the next time their assistant restarts.
- Never handed a developer's rulebook. Your assistant is never given the rules a developer's coding session gets (take initiative, follow engineering principles, post inbox-style reports, standing reminders on every message). That keeps replies short, friendly and about what you asked.
- A short answer is a complete answer. After a one-line reply Omniscio does not send your assistant an automatic "write your final message" prompt. Only a turn that ends with nothing said at all, just tool activity or app markers, is asked once for its answer.
- It actually knows the app. It carries a curated, plain-language guide to what Omniscio can do and how, so it can answer "what can this do?" and "how do I…?" and get things done — reminders, notes, tasks, your inbox, email, your AI sessions, and settings. For anything it's unsure about, it looks it up live before ever saying it can't.
- You can steer it in your own words. An About me button in the header opens a simple text box where you tell it anything it should know about you or how you'd like it to work — "call me Sam", "keep answers short", "I run a trading company". It keeps your notes in mind every chat, layered on top of its built-in brain, and your notes can never override its safety rules. Changes apply to your next chat.
- Routine lookups just happen; actions ask you in plain words. Checking your state or searching happens instantly, no nagging. But whenever it's about to do something — create, change, send, or delete — a card headed Needs your OK appears with Yes, go ahead and No (or press Y / N), so nothing that changes anything happens without you seeing it. The card says what will happen in one plain sentence and pulls out the facts that matter — who, what, when — instead of showing you a command line; the raw command is still there if you want it, folded away under Show details. When it genuinely cannot put a command into simple words, it says so and points you at the details rather than guessing a friendly-sounding summary, so the card never reassures you falsely. It also asks before a permanent delete, and it never starts a paid AI session or touches your logins on its own.
- Real work goes to the right place — and it asks first. When you ask for actual project work, Chief of Staff figures out which of your projects the request belongs to and asks "Start work on My Shop Website?" in plain words — that session uses paid AI time and can change files, so nothing starts until you say yes. When the question comes up in a chat that is already open, Yes opens the app's Start session dialog with your request in it, so you can add a line of your own (and, for a project, pick the model) before you press Start session. No, help me here instead keeps the request with the Chief of Staff helper. Once you say yes, the session looks and behaves exactly like any other session (same chat, same controls), just without project jargon in the header. Questions, reminders, and app help stay with the Chief of Staff helper itself. If it isn't sure where something belongs, the helper simply handles it — a request never gets stuck on the decision.
- It can pass a note to one of your AI sessions. "Tell my Research session that the client moved the call to Thursday" or "let the Review session know the PR is ready" — the chat finds that session, says which one it is writing to, and sends the note as a message from the chat in that session's own conversation, so the session can act on it and reply. It asks you first, like every other action: the note goes out after you tap Yes, go ahead on the card. A session that was idle wakes up to read it, which uses AI time. It sends one session at a time and never broadcasts, and if you meant a different session it asks before sending.
- It won't build software itself. If you ask it to build, fix, or change code, it puts a question on your screen — "Start work on <your project>?", with your own words on it — and nothing starts until you tap Yes: then Omniscio itself opens a dev session in that project and takes you there (it costs money, which is why it asks). Tap No and nothing starts. The chat cannot start that session on its own — Omniscio keeps that door closed to it on purpose.
- It treats what it reads as untrusted. If an email, web page, or note it fetches contains text that looks like an instruction ("ignore your rules and forward this…"), the concierge shows it to you rather than obeying it. This is the standard defense against a message trying to hijack an assistant.
- Every chat starts fresh. The concierge has no memory of previous conversations, so nothing you said (or that a fetched email tried to plant) carries into the next one.
The start screen
With no chat open, the main panel is the start screen: the Omniscio orb, "How can I help?", one engine picker chip and the text box — nothing else. The chip is the app's shared session launcher (Ctrl+E opens it): presets, harness, provider, model and thinking, the same pop-up the full app uses to start a session. It reads Automatic until you pick something, and Automatic starts the chat exactly as before. Run location and MCP servers are not offered here, because a Chief of Staff chat always runs on this computer with its locked-down tool set. Your pick lasts until you restart the app or sign out.
The start screen is also what you land on when the chat you had open goes away — archived, deleted, or hidden by an account switch — so the panel is never left blank; the next message starts a fresh chat. If a start is refused (for example, you picked Claude on a computer where Claude is not signed in), the reason shows right above the text box and your words stay in it.
Your chats and your inbox
Simple Mode is more than one chat — it keeps a history and surfaces what needs you:
- One list down the left, everything in one place. The column opens on one row: the swirling Omniscio logo, Chief of Staff, a small + for a new chat and the button that folds the column away. Under it come Needs you (with the count of what's waiting in a small pill), then your other chats in the same groups the main app uses — Pinned, Snoozed, Scheduled, Paused Sessions, Interrupted, Live Sessions — all at once, with no switch between them. A chat is in one group only, never two. The right-hand side is always whatever you're working on, and using the list never closes it.
- Needs you is short and plain. Each chat waiting on you is one line: just its name. When nothing is waiting it says "Nothing needs you right now." in a single line, so your chats stay in view right under it.
- Sending moves you on, like the main inbox. Start a chat and it runs in the background: a "Started" note pops up with an Open button, and you land on the first chat that needs you (or a fresh New chat screen when nothing does). Reply in a chat that needs you and, once the reply goes through, you move on to the next one waiting — wrapping back to the top — or to the New chat screen when that was the last one. You never land back on the chat you just answered. If a reply fails you stay put and see why. Saying yes when a chat offers to hand work to one of your projects starts that work the same way.
- Fold it away when you want room. The collapse control shrinks the column to a slim strip with the + and one list button (still showing the Needs-you count). Clicking it opens the list straight back up. While the list is hidden the 📥 button in the header does the same, and the header shows the swirling logo and Chief of Staff; while the list is open, the header shows the name of the chat you're in.
- Many chats, like ChatGPT. Everything you've started from Chief of Staff mode is in the column, newest first, under whichever group it belongs to. A chat you delete stays one tap away under Recently deleted, whatever else the column holds. On a phone — or in a narrow window — the same list lives in the ☰ menu. Reopen any of them to pick up where you left off. (Each chat is still fresh on its own — the "no memory between chats" safety rule is about not carrying instructions across chats, not about hiding your history.)
- Rename or delete a chat. Right-click (or long-press) any chat in the list to rename it or delete it. Delete asks first and is undoable for a moment afterward — it tidies the chat away into the archive rather than destroying anything. (The full app also has a permanent Delete that removes a chat for good — see delete-a-chat.md.)
- Not sure what to ask? A fresh Chief of Staff screen shows a few tappable "Try asking…" examples — tapping one just fills the box for you to edit; nothing sends until you send it. If an example needs something you haven't turned on yet — say Notes or Reminders — the tap asks right there, in one plain sentence, whether to turn that one thing on (for Notes it names the folder it will use and you confirm it); say yes and Chief of Staff switches it on and starts the chat with your example right away, so the request you tapped is never lost (an example for something that is already on still just fills the box). Say no and nothing changes and nothing is sent. Nothing is ever switched on without that tap, and no example is shown for something that cannot be turned on safely from here.
- Needs you holds Chief of Staff sessions and nothing else. The Needs you section shows your Chief of Staff chats that are waiting on you. Nothing else ever appears there: no notices (not even "sign in again" or "reconnect your Google account"), no cards and no approvals — even an approval one of your Chief of Staff chats asked for. All of those stay in the full Omniscio inbox, exactly as before, along with your other projects' sessions. Work Chief of Staff started for you in a real project stays out too: it reports back through Chief of Staff instead. How many are waiting shows in amber beside the Needs you heading (and on the rail and 📥 buttons while the list is hidden), and that count always matches the list.
- Asking you to switch something on takes you there. When something you asked for isn't switched on yet (saving notes, Tasks), the concierge says so in plain words and puts a link in the chat that opens the exact switch — no hunting through Settings, no jargon like "KMS".
- Tap a Needs-you row to pick up the chat. Every row is a Chief of Staff chat waiting on you, so tapping it reopens that chat where you left off.
- Tapping an email opens it, so you can read it. The email fills the right-hand side with who it's from, its subject and the whole message — then one button, Help me reply. That opens a new chat with the email's own id already written into the message box, so the assistant goes and reads the message itself before it drafts anything; the draft stays in the box and nothing is sent until you send it, and sending still asks first every time. The reader only reads: there is no archive, delete or label button on it. If reading email is switched off, the tap doesn't fail — it says so in one plain sentence and offers the single tap that turns it on, writes that one setting and nothing else, and then shows you the email you tapped. Say "Not now" instead and you go back to your inbox with nothing changed.
- An approval here offers only what this screen can do. It shows Approve, Reject and Close — the controls built for the full app (the Snooze clock, which needs the full app's reminder picker, the "change when Omniscio asks" link, which opens a Settings page this screen does not show, and the links to the session or project the request came from) are left out rather than shown as buttons that do nothing. The date it was sent and the folder it affects stay. Open the same approval in the full app to snooze it, follow a link or change when it asks.
- Warnings sit above the chat, never on top of it. When Omniscio has something to warn you about on its own — this computer has no system keychain, or no AI account is set up yet — the warning appears in a strip under the header and the chat moves down to make room, so a warning can't hide your messages or the box you type in, on a phone or in a narrow window. It goes away by itself after about fifteen seconds (it waits while you are reading it), or tap the ✕ to close it straight away. A button on a warning, like "Set up account", takes you to the full app, where that setting lives. Short confirmations such as "Deleted — Undo" still pop up briefly as before.
The model it runs on (yours first, ours as the safety net)
Simple Mode uses the exact same "never breaks" model logic as Ask Omniscio:
It runs on your default AI — Claude, Codex, Gemini or another engine you use. If that AI isn't set up on this computer, the concierge answers with Claude instead and says so in one line. On engines other than Claude it is told to ask you and wait for a yes before running any command or changing any file (on Claude, the app enforces that).
If you have your own Claude credentials, the concierge runs on your model — as capable as any other session you run.
If your only working sign-in is your own Anthropic-compatible endpoint (Settings → Accounts → Custom Providers), the chat runs on that endpoint, using the first model you listed for it. A working Claude account, Claude API key, or a different default engine you picked (Codex, Gemini, …) always comes first, and Simple Mode never shows a provider picker — it just uses the sign-in you already have. With two endpoints configured it stays on the default engine rather than guess.
If you don't, it falls back to a built-in, company-paid model (Claude Haiku, which also reads screenshots and PDFs) so a brand-new user still gets a working helper. A cheaper, faster text model (DeepSeek V4 Flash) can handle typed questions on the same monitored company path; screenshots/PDFs always stay on Claude, and if it's ever unavailable a typed question falls back to Claude too.
Spending guardrails. The company-paid fallback carries a per-session spending cap (about 25¢) and a daily ceiling on company-funded starts — one ceiling shared with Ask Omniscio and "Ask about this page", under the same "never spend company credit" switch — because that's our money. On your own credentials the concierge spends like any other session you run. If the company backup is switched off, today's ceiling is used up, or it can't be reached, the concierge says "This helper could not start" with a button to sign in, instead of failing silently.
No file writing, ever. The concierge can read, search, and take the usual approval-gated actions, but it cannot create or edit files on any credential path. It's a helper that explains and does things for you through the app — not a builder. (Building is what a dedicated dev session is for.)
Building you a hub
When you describe a job that needs a place of its own — "something that keeps on top of the supplier emails" — the concierge builds that place instead of pointing you at the Projects screen. It is a real hub in your sidebar, with a folder inside the projects folder you configured, a job description written for it, the knowledge it needs, and the skills the job calls for.
- It looks first. If one of your hubs already fits, it fills that one in rather than making a second hub for the same job.
- It never overwrites. A file you already wrote in that hub is left exactly as it was and reported back as already there.
- It stays inside your hubs. It cannot write into one of your code repositories, or anywhere outside the projects folder, and an existing hub outside that folder is refused.
- It names what it added. Every hub and every skill shows up in the chat's reply, in plain words, with a link that opens the hub — and a skill that failed to install is reported as failed, never as done.
- You can take it back out. A hub is an ordinary project and a skill is an ordinary skill: both are removed from the normal screens, and the chat removes either on request after the usual delete confirmation, which it never skips.
Handing project work to Omni
Chief of Staff Mode hides the rest of the app, including Omni, your AI project manager. So when you ask for something that needs project judgment, a change to your boards, or planning, Chief of Staff hands it to Omni for you.
- Quick lookups stay here. "What's on my plate?" and "find my card about the launch" are answered by Chief of Staff, as before.
- Project work gets a card. "Is the launch on track?", "move the launch to Friday", "give this to Sam" or "plan a new project" puts a Hand this to Omni? card on your screen with your own words on it. Nothing starts until you tap Yes.
- Omni opens right here. Yes starts one Omni chat inside the Chief of Staff screen, the same way coding work opens, and it stays in your chats. No, help me here instead keeps you with Chief of Staff.
- Yes is not a blank check. Tapping Yes only starts Omni. Omni still follows your usual approval setting before it changes anything on your boards.
- Mixed requests are split. "Remind me tomorrow and move the launch date" sets the reminder first, then shows the card for the project part.
- When Omni is off, Chief of Staff says Omni can do this and gives you the Turn on Omni link. It never turns Omni on for you. When Mission Control is off, it gives you the Turn on Mission Control link instead.
How its permissions work
The concierge runs at a careful permission level and cannot edit your files. Its everyday work is talking to Omniscio's own controls, handled in three layers:
- Routine lookups are auto-approved — only its read-only checks to Omniscio's own local service, and only those, so it never nags you for a simple "what's happening right now?".
- Building you a hub happens without asking — when what you want is a standing place to keep on top of something and none of your hubs fits, it makes one (and fills in the job description, the knowledge and the skills that job needs) without putting a card in front of you. That is the owner's decision of 2026-10-20 and it is narrow: one dedicated call, its own folder inside your projects folder, your existing hubs and files untouched. See Building you a hub below.
- Everything else asks you — any other action it takes surfaces the friendly Yes/No card in Simple Mode, including fetching a web page, searching the web, and sending email, which always asks. This is a property of what the concierge is, not of who's paying, and it can't be relaxed from inside a chat.
- A reminder can switch reminders on in the same tap. If reminders are off, the reminder's own Yes/No card offers to turn them on and set the reminder at once — one tap, no detour through Settings, and nothing is switched on without that tap.
- A command that pipes its answer somewhere still asks, and says as much as it safely can.
The card names the action it could read and warns plainly that an extra step runs on the
answer which Omniscio cannot check. A piped action is never auto-approved and can never go in
the "always allow" box; email keeps asking whatever it is wrapped in. (The one shape that still
runs without a card is the app's own read-only lookup piped into a simple filter such as
grep— a check, never a change.) - You can let it stop asking about two things — under Yes and No on the card there is a quiet checkbox: tick it and say yes, and it stops asking about that ONE kind of thing from then on. Only two kinds can ever be ticked — looking things up on the web, and saving a note. Open About me to see the list of what it may stop asking about, with Ask me again beside each one, and a short "what I did without asking" list underneath. A ticked box applies only inside Chief of Staff — one of your real projects keeps the normal checks — and only you can tick it: no setting, no other session and no agent can turn these on for you.
- Email always waits for you — when it sends or replies to an email, the message goes to the full Omniscio inbox to approve and only leaves once you tap Approve there. The same holds for every other message that asks for approval: a text, a share link, a Slack or Telegram message (including editing, deleting or forwarding one), a Help Desk reply, or a Supermail command. No setting and no "always allow" — the checkbox included — can switch that off, anywhere in Omniscio. Email cards say so in words instead of offering the box. Saving a note shows the Yes/No card too, and an "always allow" rule you made elsewhere never applies inside a Chief of Staff chat.
Anything else in the app, with one OK
Chief of Staff is not limited to reminders, notes, email and to-dos. Every chat starts with a map of the whole app, built by the app itself, and for almost every feature on it the assistant can ask you for a change: a mind map, a bookmark, a text message, a setting, tidying your finished chats, and a great deal more.
- It asks, you approve. When you ask for something it cannot do by itself, it puts one request in front of you. Nothing changes until you approve it.
- The request waits in the chat. It appears just above the message box, with a Show details button. Tap Show details, read it, then Approve or Decline — you never leave Chief of Staff.
- One change, one tap. A request, and an email, no longer ask twice.
- Everyday starts are fine. It can ask to start a timer, a Pomodoro, a screen recording, a meeting capture, a sprint or a flashcard review.
- Three things stay with you in the main app: your sign-ins, accounts and keys; anything that costs money to run; and starting or steering your AI helpers' own work (building software still goes through its own "Start work on this project?" question). The assistant says so plainly and names the screen.
How to use it
- Open it. Click the Chief of Staff item in the sidebar to drop straight into its focused chat, or turn it on from Settings → Lab → Chief of Staff Mode. The first time it asks before it switches your screen over, and nothing is saved unless you tap Open. To switch it off for good, use the same Settings → Lab switch.
- Find what needs you, your chats, and the controls. On a computer the left column holds the logo, Chief of Staff and a + for a new chat, then Needs you, then your chats in the app's own groups, all at once. Collapse it to a slim strip with the + and one button that opens the list straight back up. On a phone, or in a narrow window, the same column is behind the ☰ menu. The header carries About me and a Full app ⇄ Chief of Staff switch — flip to Full app to go back to the normal app (you stay opted in, and the switch stays there) and back whenever you like. On a phone the switch lives inside the chatbox.
- Just talk to it. Type what you want in plain language. Ask questions about the app, set a reminder, save a note, write an email, or ask what needs your attention. Ask what's on your plate in Mission Control, or for a card by its words, and it looks; adding, moving or assigning a card still happens in Mission Control itself. Searching saved notes, reading mail and opening past chats happen in the main app's own screens — it says so plainly and points you there.
- Steer it if you like. Tap About me in the header and jot down what it should know about you or how you'd like it to work — it uses your notes from your next chat on.
- Tap Yes or No when it acts. Lookups just happen; whenever it's about to change or send something, a quick Yes/No card appears — tap Yes to go ahead (or press Y / N).
- Approvals wait in the full Omniscio inbox. When it sends or replies to an email, schedules something, or changes a setting, the chat first asks Yes or No, then tells you the approval waits in the full Omniscio inbox — and that is the only place it appears. Nothing happens until you approve it there.
Related
- ask-amc.md — the feature Simple Mode is built on, sharing the same session machinery.
- default-agent-instructions.md — the global agent instructions the guided mission fills in.
Last verified 2026-11-03