Google Analytics, Vercel Web Analytics, Leadsy and the TikTok Pixel stay off until you accept. The help chat starts once you have chosen. The scheduling button starts once you have chosen. Read our Privacy Policy.

Skip to content
StellarFirmStellarFirm

Orbit

Coming soon

Proactive AI Agents that check in before you ask.

Orbit is a team of proactive AI assistants. Pick an assistant, write it a brief, choose the services it watches and how often it checks, as often as every hour. It looks only at what is new, does what the brief asks and tells you when something needs you.

A check only reads, and anything that sends, posts, spends or deploys still waits for your Approve.

In your Orbit

Three assistants that check in for you.

Each one watches the services you connect and does what its brief says. A check only reads, so nothing is sent, posted, spent or deployed on its own.

The Coder, on call

Your AI SRE agent: an AI on-call engineer that checks as often as every hour, reads your Vercel logs, Better Stack monitors and incident.io alerts, and tells you what it found.

Watches

  • GitHub: pull requests, failing checks, reviews, issues
  • Linear
  • Vercel: deployments, build and runtime logs
  • Better Stack: uptime monitors, incidents, logs
  • incident.io: incidents and alerts
  • Sentry
  • Datadog
  • PagerDuty
  • UptimeRobot

What it does

It finds the failing build or the new errors, reads the logs, drafts the issue and a fix plan, and tells you.

On a check it only reads. Ask it in its chat to fix the problem, and pushing the fix waits for your Approve.

The Executive assistant

An AI executive assistant that has your day ready before you open it.

Watches

  • Gmail or Outlook
  • Google Calendar
  • Slack
  • Google Drive

What it does

A morning brief. Replies drafted for your Send. Calendar conflicts spotted. Meetings prepared.

It never sends a reply or changes an invite on its own. A draft waits on its card until you press Send.

Marketing

An AI marketing agent that notices what is launching and drafts the work.

Watches

  • Slack: channels
  • Google Drive: folders
  • Gmail or Outlook: your inbox

What it does

It drafts the launch: posts, emails, campaigns and ads.

Launching ads spends your money, so it always waits for your Approve in the chat. A check never launches or posts anything.

Models

Any model, on every assistant.

Each assistant runs on the model you choose for it, or on Auto, which picks for each task. Claude, GPT, Gemini and Grok come with a paid plan, open models work on every plan, and you can use your own key too. The models you set for an assistant also run its checks.

  • Auto
  • Claude
  • GPT
  • Gemini
  • Grok
  • Open models
  • Your own key

How it works

Set it up once. It checks in from then on.

  1. Turn on Works on its own

    On an assistant's page, when you hire one, or by asking the CEO assistant. The assistant joins your Orbit.

  2. Give it a brief and a schedule

    Write what it should do, pick the connected services it watches and how often it checks, as often as every hour.

  3. It checks and reports

    Each check looks only at what is new and answers in the assistant's updates chat. A check with nothing new is one quiet line there. When it finds something, the bell lists it, and a step that waits for your Approve also reaches you by push on your phone and by the desktop app.

  4. You approve what goes out

    A check only reads. A draft waits on its card for your Send, and anything that sends, posts, spends or deploys waits for your Approve.

Checks that keep going.

  • Picks up where it left off

    A check that a restart cuts off carries on from where it stopped.

  • Waits as long as you need

    A check that waits for your Approve is still waiting when you come back.

How Orbit compares

A Claude Duet and Grok alternative: a team, on any model.

Claude Duet is one proactive personal assistant, on Claude. Grok's teammates are agents you set up yourself, on one shared cloud computer. Orbit is a team of ready-made assistants with roles, each on the model you choose, and its checks only read.

Orbit compared with Claude Duet and Grok's teammates
FeatureOrbitClaude DuetGrok teammates
What it isA team of AI Agent Assistants with roles, led by the CEO assistantOne proactive personal assistantAI teammates you create, each with a name and a job you describe
RolesReady-made: the Coder on call, the Executive assistant, Marketing and more, each with its own skillsOne general assistantNone ready-made: you describe each job, with example prompts to start from
ModelsAny model on every assistant: Claude, GPT, Gemini, Grok or open models, or Auto to pick for each taskClaudexAI's docs name no model and no way to choose one
When it worksOn the schedule you set, as often as every hour, looking only at what is newAlways on, with a brief every morning, reaching out when it judges something needs youAlways on, with routines you teach it by showing it, run on a schedule or after an event
What it watchesThe services you pick: GitHub, Linear, Vercel logs, Better Stack, incident.io, Sentry, Datadog, PagerDuty, UptimeRobot, Gmail or Outlook, Google Calendar, Slack and Google DriveApps you connect, from more than 6,000 connectors and pluginsThe apps it signs in to, with connectors where there are some and the browser for the rest
How it reaches themRead-only calls through the connections you made. A check never opens web pages or uses a computerIts own cloud computer, using the apps you connectOne cloud computer with a browser, files and a terminal, shared by all your teammates, logins included
ControlA check only reads. Anything that sends, posts, spends or deploys waits for your Approve, and money always asksAsks before sensitive operations like sending messages or payingSensitive or consequential actions can stop for approval, depending on the tool, the risk and your Auto-review rules

Like Grok's models? Pick Grok for any Orbit assistant, next to Claude, GPT, Gemini and open models.

Claude Duet details are from claude.com/product/claude-duet, and Grok's from xAI's launch post and documentation, as of 27 October 2026. Claude is a trademark of Anthropic. Grok is a trademark of xAI.

Questions

What is Orbit?

Orbit is StellarFirm's name for AI Agent Assistants that work on their own. You turn on Works on its own for an assistant, give it a brief, pick the connected services it watches and say how often it checks. On that schedule it looks at what is new, does what the brief asks and tells you when something needs you. Each assistant you turn it on for joins your Orbit.

What can a check do, and what can it not do?

A check only reads. It looks at what is new in the services you picked, does what your brief asks, such as summarising, triage or planning, and answers in the assistant's updates chat. Where the assistant can write a draft, the draft waits on its card for your Send; where it cannot, it puts what it would send in its answer.

A check never sends, posts, spends, publishes or deploys, and it changes nothing in your services, so it cannot acknowledge, page or resolve an incident. It does not open web pages, search the web or use the assistant's computer.

What you ask the assistant to do afterwards in its chat follows your approval rules: anything that sends, posts, spends or deploys waits for your Approve unless you switched on auto-approve for that kind of step, and money always asks.

Which services can an assistant watch?

Gmail, Outlook, Slack, Google Calendar, Google Drive, GitHub and Linear, and for production work Vercel (deployments, build and runtime logs), Better Stack (uptime monitors, incidents, logs), incident.io (incidents and alerts), Sentry, Datadog, PagerDuty and UptimeRobot. A service has to be connected on Integrations first, and the assistant's role has to be able to read it. GitHub connected only with a token reads issues, not pull requests or failing checks, and the setup form says so beside it.

Which models can the assistants use?

Any model, on every assistant. Each assistant runs on the model you choose for it, or on Auto, which picks for each task, and its checks run on the same setup. Claude, GPT, Gemini and Grok come with a paid plan, open models work on every plan, and you can use your own key too.

Where does Orbit run?

Checks run in StellarFirm cloud, so they go on while your computer is off. You set Works on its own up on StellarFirm web, and the updates appear in the assistant's chat there. The bell lists what a check found, and a step that waits for your Approve also reaches you by push on your phone and in your browsers and by the desktop app. Each of those has its own switch in Settings.

What does it cost?

Each check uses credits, like any message you send an assistant, so the more often it checks, the more it uses. You can set a spend limit for the company or for one assistant in Settings. When a limit is used up, new work pauses until it resets or you raise it.

Is Orbit an alternative to Claude Duet or Grok?

Yes. Claude Duet is one proactive personal assistant, on Claude, working on its own cloud computer. Grok's teammates are agents you set up yourself, and they share one cloud computer with its files and logins. Orbit is a team of ready-made assistants with roles, each on the model you choose, Grok included, and a check only reads through the connections you made: anything that sends, posts, spends or deploys waits for your Approve. Claude Duet details are from claude.com/product/claude-duet, and Grok's from xAI's launch post and documentation, as of 27 October 2026.

When can I use Orbit?

Orbit is coming soon and is not open yet. Join the waitlist from this page: joining reserves a place for a later invite and charges nothing.

Be first when Orbit opens.

Join the waitlist. You call the shots. Your AI Agents make it happen.