OmniPotem
Waiting list

Coming soon

OmniPotem

Run several coding agents in parallel on your Macs, without staying at your desk, and approve every push after reviewing it.

Not another agent: OmniPotem starts Claude Code and Codex as they are, with your subscriptions, in the repository you assign to them.

Join the waiting listSee the features

The service is not open to the public yet.

Why

Why OmniPotem

OmniPotem is a web cockpit, installable on your phone, for developers who run coding agents (Claude Code, Codex) on their Macs. Follow every session, answer the agents’ questions, review their work before the push and get their reports.

Not another agent: Claude Code and Codex, as they are

OmniPotem does not reimplement an agent. Its daemon, which calls no model, starts the agents published by Anthropic and OpenAI, Claude Code and Codex, with the command set on your Mac and your own subscriptions: their updates apply as they are. Claude Code and Codex today, and soon other coding agents.

In the right place: each agent starts in the folder set for it on the Mac, for example its project’s repository, just as if you had started it by hand in a terminal. There it reads the project’s instructions (CLAUDE.md, AGENTS.md) and has the machine’s tools and access. Delegated work starts in the folder of the agent that receives it.

Claude or Codex, by quota and by task

The daemon reads each Mac’s Claude Code and Codex quotas (5-hour and weekly windows) every 5 minutes. At 80% and 90%, an alert reaches the lead agent, the one that starts and follows your delegated work. Its picker recommends the other agent when one is past 93% of its week or 99% of its 5-hour window, if the other has room.

For delegated work, the lead agent picks the Mac and the agent with each machine’s remaining quotas, load and memory in view; splitting work by kind of task follows the instructions you give it. Nothing switches on its own: you change agent in one tap, or the lead agent decides.

Without OmniPotem, with OmniPotem

  • Without: Each agent is started by hand, in the right folder, on each Mac.

    With: The daemon starts Claude Code or Codex in the folder set for each agent, on the Mac you choose, from the cockpit. See

  • Without: One terminal per agent, to check one by one.

    With: Every session’s thread and the state of every delegated work, with its Mac, repository, model and effort, in one cockpit. See

  • Without: An agent’s question waits until you are back at your desk.

    With: It arrives as a notification on your phone; you answer in one tap, with five seconds to cancel. See

  • Without: Each diff is read in the terminal, without knowing why each file changes.

    With: A review with an overview, a reason per file and the tests run; only the commits you approve go out. See

  • Without: Several approved branches each go out on their own.

    With: Releases gathers a repository’s approved reviews into one deployment, in their stack’s order. See

  • Without: You find out about the quota limit halfway through a piece of work.

    With: Each Mac’s quotas are read every 5 minutes, with alerts at 80% and 90%, and a recommended agent. See

  • Without: Nothing shows where each repository stands.

    With: Where each repository stands, kept up to date by the lead agent, the state of the sessions and what waits for you, on one page. See

  • Without: An agent’s write-up stays in its terminal, then gets lost.

    With: The lead agent publishes its reports in the cockpit, readable as a page or a presentation, with their earlier versions. See

Follow

All your repositories and Macs on one page

The lead agent starts delegated work, each piece in its own session on one of your Macs, and keeps track of each repository. You see who is working, where, with which model and effort, and what waits for you.

The repository cockpit

What it does
The home screen brings together where each repository stands (live, under way, what waits for you), tasks by project, the state of the lead agent and of each Mac, the most urgent card, and the delegated work and sub-agents at work.
How to use it
The lead agent keeps each repository’s status up to date. The “For you” column gathers questions, reviews and cards, with “Next”; the Accomplishments page sums up the week’s pushed reviews and closed tasks.
Example
The api repository reads “Waits for you: Your OAuth 2.1 review, then the staging migration”, and the review goes to the top of “For you”.
The list of delegated work: five pieces on two Macs and four repositories, two at work, one waiting for a decision on checkout.spec.ts, one finished today.

Delegated work

What it does
The lead agent starts a piece of work in a separate session, on one of your Macs, in the folder of the agent that receives it. The cockpit shows its state, summary, agent with model and effort, background tasks and latest action live, and keeps its report; its sub-agents appear under it.
How to use it
On a task, hand it to the lead agent, with an optional note: it picks the Mac and the agent and starts the work. The work’s questions are answered on its page, its thread stays read-only, and “Interrupt” sends it a single Escape key.
Example
“fix-flaky-checkout-e2e” on the MacBook Pro: working, then waiting for a decision on checkout.spec.ts, then done with its report.
A report from the lead agent: api’s CI goes from 12 min 40 s to 5 min 10 s, version 2, with its summary.

The lead agent’s reports, as pages or presentations

What it does
The lead agent publishes its reports in the cockpit: headings, tables, key figures, step timelines, “Key point” or “Decision” callouts, diagrams and charts, secrets masked. Publishing a report again keeps its 10 previous versions.
How to use it
Open Reports, filter by project, pick a version. “Present” shows the report as slides, one section each, full screen if you like: arrows or swipe to move on, Escape to go back.
Example
“api’s CI: from 12 min 40 s to 5 min 10 s”, version 2: the figures before and after, the causes, then what is left to do.
Claude Code’s and Codex’s quotas on the MacBook Pro, window by window, with the reset time and the declared subscription.

Quotas, usage and machine health

What it does
A gauge shows the Claude Code and Codex quotas (5-hour and weekly windows), orange from 80% and red above 90%; an alert reaches the lead agent at those thresholds. The Usage page brings together the month’s cost and the tokens by model, Mac, project, agent or delegated work; the Performance page, each Mac’s load, memory and disk.
How to use it
The daemon reads the percentages and reset times on the Mac every 5 minutes, with no token and no API call. The lead agent’s picker recommends the other agent when one is past 93% of its week or 99% of its 5-hour window, if the other has room.
Example
If Claude Code reaches 95% of its week while Codex is at 18%, the lead agent’s picker recommends Codex, and you switch in one tap.

Product

What OmniPotem is

Four pieces that work together: the cockpit, its phone app, a daemon on each Mac, and the coding agents that the daemon connects to the cockpit.

  1. The web cockpit

    Where everything is followed: where each repository stands, the agents’ questions, reviews before push, releases, delegated work, reports, session threads and your machines.

  2. The phone app

    The same cockpit, installed on your home screen as a web app. It notifies you of a new question, a review to open or a “To see” card.

  3. The daemon, on each Mac

    A Node program, with no AI model, that runs on your Mac in cmux. It relays your messages to the sessions, sends their threads to the cockpit, follows delegated work and computes the diffs of reviews.

  4. The coding agents

    Claude Code and Codex, and soon other coding agents, started in cmux sessions on your Macs, with your own subscriptions. They talk to the cockpit through MCP tools: ask a question, submit a review, deliver a task. The lead agent starts and follows delegated work.

Who it is for

Developers who run several coding agents, on one Mac or more, and want to follow them and make decisions without staying at their desk.

Requirements

macOS, cmux, git, Node 22 or later, and Claude Code or Codex. FileVault turned on, required by default.

How it works

How it works

The path of a piece of work, from the machine to the report.

  1. 1Add a MacIn the cockpit, you add a machine: its token is shown only once. On the Mac, the setup script stores it in the Keychain, and the machine shows up online.
  2. 2Start an agent or hand over workYou start a session from the cockpit, write to an agent, or hand a task to the lead agent, which carries it out as delegated work in a session of its own.
  3. 3FollowEach session’s thread shows the messages and one line per action. Delegated work shows its state, its model, its effort and its sub-agents.
  4. 4DecideA question or a go card arrives with a notification. You answer in one tap; the agent is woken and reads your answer.
  5. 5Review before the pushThe agent submits a review; the Mac computes and masks the diff and sends it. You allow the push of those commits or request changes. The push guard blocks any push that was not allowed, on the repositories where it is installed.
  6. 6Get the reportThe agent delivers the task with a summary, its proofs, its checks and the risk. The lead agent publishes its reports in the cockpit, and the Accomplishments page sums up the week.

The service opens first to the people on the waiting list. An address is enough.

Join the waiting list

Features

The features in detail

For each feature: what it does, how to use it, and an example. They come in four parts: follow, decide, review, steer.

The captures and examples use made-up data.

Decide

Answer the agents from your phone

When an agent reaches a decision (run a migration, quarantine a test, deploy to staging), it asks you from its session, with its choices and its recommendation. The notification reaches your phone; one tap answers it.

  • One tap, five seconds to cancelThe answer leaves after a 5-second countdown, then the page follows its delivery: received by the Mac, agent woken up, answer read.
  • Urgent firstAn urgent question goes to the top of the list, with a red edge and a notification marked “Urgent”.
  • A go for one precise actionBefore a sensitive action such as a deployment, the agent asks for a go: the button sends exactly the text shown to its session, in your name.
A question from the lead agent on a phone, right after a tap: “Shall I run the migration on the staging database?”, the recommended choice selected and the answer leaving in 4 seconds, with a link to cancel.

Questions, go cards and “To see” cards

What it does
An agent asks a question with its choices (twelve at most), an urgency and a due date. A go card asks for your approval of one specific action and shows the exact text that will be sent. A “To see” card points out something to look at, with two reminders at most, never at night.
How to use it
You answer in one tap, with five seconds to cancel and an optional note, or you discuss it with the agent under the card first. A go card is declined with “Not now”. Answering requires a two-factor code from the last 7 days, giving a go one from the last 12 hours.
Example
Go card: “go deploy web staging”. The button sends that text to the agent’s session, unchanged and in your name.

The phone app and its notifications

What it does
The cockpit installs on your home screen and sends notifications: a new question (marked “Urgent” when it is), a review request, a “To see” card and its reminders, an agent’s reply in a discussion, prose ready, a second review returned.
How to use it
Add the cockpit to your home screen (required on iPhone), then turn notifications on with the bell, on the Decide page or in the settings. Tapping a notification opens the question or card it is about.
Example
On the train, the notification “Urgent: Shall I run the migration on the staging database?” takes you straight to the question to answer.
The overview of the review “Migrate authentication to OAuth 2.1 with PKCE”: before, the access token travels in the URL; after, authorization code with PKCE and rotating refresh tokens.

Review

Review the work before the push

The agent submits a review of its branch: what changes, why, the tests it ran and its limits. The diff is computed and masked on your Mac. You read the overview, then each file with its explanation.

  • Before, after, in figuresThe overview sums up the change and gives its key figures, before the first file.
  • A reason for each fileEach component touched states its role and its risks; each file, why it changes.
  • Only approved commits go outOn each repository where you install the push guard (a pre-push hook), only approved commits go out. A review built on another’s branch is read in the stack’s order.

Review before the push

What it does
The agent submits a review: a summary, one sentence per file, the tests it ran and the limits. The page shows an overview (what is touched, user journeys, risks), the diff side by side or unified, and your “Seen” marks. A review built on another’s branch opens in its stack’s order.
How to use it
You comment on a line, ask the agent about a line or discuss the review before deciding, then go over your comments one by one. You allow the push of those exact commits, for 7 days, or request changes. Opening the code and deciding require a code from the last 12 hours.
Example
“Migrate authentication to OAuth 2.1 with PKCE”, 9 files: you mark each file as seen, ask why mobile clients before 3.2 keep the old flow, then allow the push.

Prose translation and a second agent’s review

What it does
The authoring agent rewrites a file in plain prose for readers who do not read code, on its own or annotated beside the code. A review, adversarial or simple, can be handed to a model other than the author, which returns a verdict (nothing blocking, to fix, to rethink) and comments on lines.
How to use it
“Translate into prose” on one file, or “Translate everything into prose”; “Ask for a review”, choosing the model and the effort. The prose is written in French for now, kept by the service, never in the repository, and erased with the code.
Example
An adversarial review by a Codex model: verdict “To fix”, with two comments on a race condition in the token refresh.

Releases: one deployment for several reviews

What it does
A release gathers the approved reviews at the head of a stack, in the stack’s order: in preparation, ready, partial, then live. Its page shows the reviews, the deployment, the production check, the history and how to go back to the previous release.
How to use it
From “For you” or a stack’s card, open the release, then “Go live” (a code from the last 12 hours). The lead agent checks every approval, pushes the top of the stack for a single deployment, checks production and attaches its proofs. Nothing runs from the page itself.
Example
“Release of 15 Oct, no. 2” on web: three approved reviews (OAuth 2.1, rate limit, checkout fix) go out in a single deployment.

Steer

Stay in control of the sessions and the Macs

Write to an agent from its thread, interrupt it, take an instruction back, send a message to several Macs, share a machine: all from the cockpit, on a computer or a phone.

Writing to the agents

What it does
Each session has its thread: your messages, the agent’s replies and one line per action, without command arguments or output. From there you write, attach up to 20 images, quote one of the agent’s messages or start a session.
How to use it
Open an agent’s thread, write, and paste a screenshot if needed: the daemon pastes your message into the session as soon as the prompt is free. Writing to an agent requires a two-factor code from the last 12 hours.
Example
From your phone: “Add the screenshot of the payment form to the review”, with the screenshot attached.

Staying in control of sessions

What it does
Session threads are kept for 30 days. “Interrupt” sends a single Escape key to a busy agent, never Ctrl+C. For the lead agent’s long sessions and each Mac’s permanent session, the daemon runs a handover: beyond 400,000 context tokens or after two compactions, the agent writes a handover note, and a fresh session picks it up.
How to use it
“Interrupt” sits in the thread, with five seconds to cancel; “Edit again” puts one of your messages back in the field so you can correct and resend it.
Example
An agent heads for the wrong branch: “Interrupt”, then “Edit again” on your instruction to name feat/oauth-pkce.

Several Macs, and guests

What it does
Each Mac has its token, its agents and its permanent session. An @machine mention sends a message to one or more Macs, five at most. You can invite other people to a machine, with a role.
How to use it
Settings › Cockpit: add a Mac, create an agent’s token, set the mention, invite. An invitation works once and lasts 7 days; roles (observer, operator, pilot, responder) can be limited to some agents and to an end date, and guests’ actions are logged.
Example
“@laptop @studio run api’s tests” goes to both Macs.

The service opens first to the people on the waiting list. An address is enough.

Join the waiting list

Your data

What stays on your Macs, what the service receives

The agents run on your machines, with your subscriptions. The service receives the sessions’ text, the agents’ state, figures from the Mac and the reviews’ diffs, not your repository.

Your Macs

The daemon, with no AI model, in the cmux terminal

  • Your repositories, branches and code
  • Claude Code and Codex, with your subscriptions
  • Tokens, in the macOS Keychain
  • The arguments and results of the commands the agents run, and the handover notes
  • Sensitive files (.env, keys), never sent with a review

The feed, the agents’ state, figures and the reviews’ diffs

The service

The online cockpit

  • The sessions’ text (messages and one line per action, without command arguments), secrets masked, kept 30 days
  • The agents’ state, and figures: quotas, Mac load, tokens per model
  • Each review’s diff, erased 7 days after the decision
  • Images attached to messages, kept 7, 30 or 365 days, or with no limit, as you choose

Questions, reviews, reports and notifications

You

Computer and phone

  • The cockpit in the browser
  • The app on the home screen, with its notifications

The service never sends a shell command to a Mac: only known kinds of commands, which the daemon runs according to your configuration.

Frequently asked questions

Which systems does OmniPotem run on?

The cockpit opens in a browser, on a computer or a phone. The daemon runs on macOS, in cmux, with git, Node 22 or later, FileVault on, and Claude Code or Codex.

Do I need my own Claude Code or Codex subscription?

Yes. The agents run on your Macs, with your accounts; the daemon uses no AI model. Today these are Claude Code and Codex, and soon other coding agents.

Is my code sent to the service?

Only a review’s diff, sent masked when the agent submits it and erased 7 days after the decision. Session text is sent too, with secrets masked, without file contents or command output. The repository stays on your Mac.

What happens if the service stops responding?

The agents keep working on your Macs. The push guard then refuses pushes that were not allowed, unless you configured it otherwise.

How is access protected?

With two-factor authentication: writing to a session, giving a go, opening or deciding a review and creating a Mac token require a code entered within the last 12 hours; answering a question, within the last 7 days. “Sign out everywhere” closes all your sessions, this one included.

Can I share a machine?

Yes, through a single-use invitation valid for 7 days, with a role (observer, operator, pilot, responder) limited to some agents. You can remove access at any time.

Which languages are available?

The interface is in English and French. It follows your browser’s language and can be switched at the top of this page or in the settings. Prose translations of reviews are written in French for now.

When does the service open?

No date has been announced. Everyone on the waiting list receives a message when it opens.

Waiting list

The service opens first to the people on the waiting list. An address is enough.

Your address is used only to tell you when the service opens. To have it deleted, write to contact@omnipotem.com.