A chief of staff that runs your agents.
OTTO remembers your people, keeps track of your promises and briefs you every morning. When work needs hands, it runs Claude Code, Codex and OpenCode on your own machines, and only asks you when it matters.
Sign-ups open soon. Follow along on Discord.
Talk to it on Discord, Telegram, email, the phone or your desktop.
The OTTO portal's Today page: what needs you, the morning brief, today's numbers and the active runs.
- Pull requestsGitHub
- IssuesLinear
- Error triageSentry
- Logs and costsAWS
- Cloud sandboxesE2B
- Run threadsDiscord
- Remote toolsMCP
- Any modelOpenRouter
- Any APIBring your own
- Email triageGmail
- SchedulingGoogle Calendar
- FilesGoogle Drive
- Notes and docsNotion
- Phone callsTwilio
- VoiceElevenLabs
- Smart homeHome Assistant
- On-chain readsEVM RPC
One conversation, everywhere.
Message OTTO on Discord or Telegram, email it, call it, or talk to it on your desktop. It's one conversation and one memory, so what you said on a call this morning is there in your DMs tonight.
- A brief at 08:00, a wrap-up at night and a review every Friday.
- It tracks what you promised and what you're owed.
- It interrupts you only when it matters, by DM or by phone.
A conversation with OTTO across Telegram, a phone call and Discord: it moves a meeting, remembers a promise and follows up.
It knows your world.
People, promises, facts and the story of your week. OTTO keeps them for you, and they stay yours alone, private even from your workspace admins.
Examples of what OTTO keeps: a person with their preferences, your open commitments, and the morning brief.
Watch from anywhere. Take over in one click.
Every agent session runs on its own machine. Open it from the portal, the desktop app or your phone, send a note, answer a permission prompt, or take over the terminal and hand it back.
An agent's terminal on vps-1 running tests, and the same session on a phone, where a permission prompt waits for you.
Scripts when it's routine. Models when it isn't.
A fixed procedure runs as a tested script you approve, not as a prompt. If the script fails, a model can repair the input or the script, inside the same budgets and approvals.
- OTTO notices repeated work and offers to codify it.
- Python and Node scripts are versioned, tested in a sandbox and approved by a person.
- Every run shows whether a model stepped in, and what it cost.
OTTO offers to turn a weekly request into a script, and the script's runs show when a model stepped in.
Built to be trusted with your day.
The rules that keep you safe live on OTTO's servers, not in a prompt.
Private by default
Your OTTO is yours alone. Workspace owners and admins can't read your conversation, people or memory.
Budgets and a kill switch
Daily and per-run budgets are hard limits. One switch stops all agent work in your workspace.
Secrets stay out of chat
Keys go in through one-time links or signed-in forms, are encrypted with their own key and never reach a prompt.
Your subscriptions stay home
Claude and ChatGPT logins run only on the machine they're on. They never go to the cloud or a sandbox.
Voice you can interrupt
Call OTTO or talk to it on your desktop, and cut in any time. Approvals on a call need your voice PIN.
Test environments that clean up
An agent reserves a test account, tags everything it creates and tears it all down when it's done.