Presence
A face that changes with what’s happening.
Sleeping, thinking, working, done. You know where your agent is just by glancing at the corner of your screen, even over full-screen apps.
For Claude Code, Codex, Cursor and Gemini CLI
Olo lives in a corner of your screen. It shows you what your agent is doing, asks before anything important (on your phone too) and remembers what you teach it.
![]()
Olomni is the presence, approval and memory layer for AI coding agents. It runs on Mac and Windows, installs with one command and is free during the beta.
Works with
Your agent already works for you. The problem is you can’t see it, it asks when you’ve stopped looking, and tomorrow it has forgotten everything.
34 seconds
From install to your first “yes” from the phone. No marketing cuts: this is the real Olo.
Olo grows with you
The face of your AI team. It sleeps, thinks, works and raises its hand when it needs you. And it grows with you: every task your AIs finish adds up, levels it up and earns it something to wear.
A day with your agent
No notifications, no noise. It just sits there, in its corner.
While your agent moves forward, you know at a glance. Without opening a window.
Publish, delete, deploy: it asks you, right here, and waits for your answer.
What you teach one agent, the next one knows. Even if you switch tools.
Presence
Sleeping, thinking, working, done. You know where your agent is just by glancing at the corner of your screen, even over full-screen apps.
Any agent
It works with the agents you already have and connects them for you. Olomni doesn’t replace your agent: it lends it a body.
Shared memory
Your AIs save what’s worth remembering. You see every item and confirm, fix or delete it. Nothing is stored behind your back.
Approvals
Nothing important happens without your yes, on desktop or on your phone. And anything that could get expensive isn’t even asked: it’s blocked.
Agent handoff
When an AI finishes, Olo notes where it left off. Open Codex in the project Claude Code just worked on, and Codex already knows. The handoff lives on your computer, with keys and tokens redacted.
Use cases
Olomni doesn’t change your tools. It changes what happens while they work.
Long Claude Code or Codex tasks without babysitting the terminal. If it needs your permission, it reaches your phone and you decide with one tap, wherever you are.
Claude Code asks · on your phoneClaude Code on the backend, Cursor on the frontend: each session gets its own lane and Olo points at the one waiting for you. Subagents show up as mini-Olos.
2 working · 1 waiting for youYou hit one tool’s limit or want a second opinion: the next AI starts knowing where the project was left.
Handoff: Claude Code → CodexForce pushes, deploys, deleting data: anything irreversible waits for your yes. If you trust something, choose “don’t ask again in this project”.
git push --force · high riskOlo remembers what your AIs finished today. Copy the summary for your standup or share it.
Today: 7 tasks doneYou decide
Your agent can do a lot. Olomni makes sure the irreversible goes through you first. Even when you’re away from your computer: the question reaches your phone.
Try it
Your agent is waiting. Your call.
These aren’t even asked
Blocked automatically, even if the agent says it’s no big deal. And Olomni never runs anything on its own: it only carries your decision.
Comparison
| Situation | Without Olomni | With Olomni |
|---|---|---|
| Knowing what your agent is doing | Going back to the terminal to check. | A glance at the corner of your screen. |
| Approvals while you’re away | The session sits waiting for hours. | It reaches your phone; you decide with one tap. |
| Dangerous actions | Approving everything blindly, or --dangerously-skip-permissions. | The irreversible needs your yes; moving money is blocked. |
| Switching from Claude Code to Codex | Explaining everything again. | It starts knowing where the project stands. |
| Several AIs at once | Five windows and guessing which one is waiting. | One lane per session; Olo points at the one that needs you. |
| Memory | A different instructions file per tool. | One shared memory you can see, fix and delete. |
Under the hood
Olo listens to your agent through its own hooks and through MCP, the open standard for AI tools. Your code never leaves your computer.
01 · your agent
Reports each step through its hooks and asks for permission over MCP.
02 · your computer
Shows the status, stops the irreversible and keeps the handoff between AIs.
03 · your account
Carries approvals and status to your devices and stores the shared memory.
04 · your phone
Pings you and you decide with one tap. No app store.
Get started
One command, no keys to copy. Olo installs, pairs with your account and connects to your agents by itself.
Step 1
Create your free account, open Devices and tap Generate my command. It’s yours alone and works once.
Step 2
On Mac open Terminal (⌘ + space, “Terminal”). On Windows open PowerShell (Windows key, “PowerShell”). Paste, Enter, and wait a few seconds.
Step 3
It appears in a corner of your screen, connected to Claude Code, Codex, Cursor or Gemini CLI if you have them. Open a new session of your agent and watch it work.
Step 1
In your phone’s browser: Safari on iPhone, Chrome on Android. Either works.
Step 2
On iPhone: Share, then Add to Home Screen. On Android: ⋮ menu, then Install app. That’s how notifications reach you.
Step 3
Type a code from your profile. When your agent asks for permission, you get a notification. Sensitive actions take two taps, never by accident.
Any agent that speaks MCP over HTTP can use Olo. In Olo’s panel, Other agent or manual setup gives you these same lines with your key already filled in, ready to copy.
Private beta · 100 founders
An accessory that will never come back, your founder number on Olo and on your card, and a direct line to the person building it. Free during the beta.
Olomni is the presence, approval and memory layer for AI coding agents. Its desktop app, Olo, shows what your agent is doing, asks for your permission before irreversible actions (even from your phone) and keeps a memory shared by all your AIs.
Claude Code, Codex, Cursor and Gemini CLI have full integration: live status, approvals and handoff. VS Code with Copilot, Windsurf, OpenCode and Cline connect over MCP, as does any agent that speaks MCP over HTTP. Olomni doesn’t replace your agent: it lends it a body.
Open olomni.com/remote on your iPhone or Android, add it to your home screen and pair it with a code from your profile. When Claude Code, or any connected agent, asks for permission, you get a notification and decide with one tap. Sensitive actions take two taps, never by accident.
Never. Olo shows what your agent does and carries your answer back when it asks for permission. Your agent does the work, and nothing important happens without your yes. Moving money, trading stocks and changing access permissions are blocked without asking.
In two ways. Handoff: when an AI finishes a task, Olo notes where it left off and the next one you open in that project gets it on start; it lives only on your computer. Memory: what your AIs save, like “this project uses pnpm”, all of them can read, and you confirm, fix or delete it.
Olo notices right away, saves where the work was left and offers to continue with Codex, Gemini CLI or Cursor with the full context. On Mac it opens the other AI in the project folder with the handoff already typed; it never switches AIs unless you choose to.
The permission card doesn’t trust the AI’s own summary: Olo checks on your computer what would really happen, such as “would erase 3 commits on origin/main you don’t have locally” or how many files a reset --hard would lose. You decide with facts, from your phone too.
Your prompts, your code and your AI’s responses stay on your computer. Only what’s needed to notify you goes to your account: each session’s status, a numbers-only summary when done, permission requests, and memories, which you see and control.
No. You paste one single-use command and Olo installs, pairs with your account and connects to your agents by itself. If your agent isn’t on the list, Olo gives you the setup ready to copy.
Yes, on Windows 10 and 11: in Devices choose Windows, copy your command and paste it into PowerShell. Linux is on the way. Phone approvals work on any iPhone or Android.
iPhone and Android don’t let an app float over other apps. So on your phone Olo is a remote: it tells you when your agent needs you and you decide with one tap.
Go to your profile and tap Revoke on that device. It stops working instantly. Then pair the new one with another code.
It doesn’t happen with the install command: Mac and Windows only block what you download with a browser. If you downloaded it by hand, try to open it, then go to System Settings → Privacy & Security → Open Anyway. Devices has the step-by-step.
Nothing during the beta: no card, on Mac and Windows. The first 100 also get the Founder laurel for Olo.
No card. For Mac and Windows; Linux very soon. Approvals from any phone.