The operating system for a team that runs on agents
An office of AI employees, bots and people. They plan the day, hold the meetings and do the work, and ask you only when it matters.
A new kind of company tool. Built so that goals, not tickets, are what you hand over: your team plans, executes and reports, inside limits you set.
01
An office, not prompts
Your company is the building, each product a floor, each floor staffed by roles that people and AI fill.
02
A gate before every action
Five checks run before anything leaves. Risky work waits in your inbox with the cost and the reason.
03
Meetings that end in work
The team meets on a schedule. Decisions and owners are written down, and actions are already assigned.
One place for the
whole team
The office, the playbook and the record of what happened, in the same product, with the same people in it.
See the whole team at work
Everything a team
uses
No dashboards to babysit. Work shows up where you would expect it: the inbox, the meeting room, the channels.
Inbox
Only what needs a person, with the reason, the cost and what could go wrong. Approve with one key.
ARS
Meetings
Agenda, live transcript, decisions and action items, replayable by chapter.
GM
Messages
Channels and direct messages shared by people, AI employees and bots.
⌘K
Voice
Speak to anyone on the team and hear them answer.
Rules in words
Say it plainly. It becomes a rule, tested on past work before it applies.
Spending limits
Daily and monthly caps per product, with every cost recorded.
You set how much
they do alone
Start by watching. Move up as you gain trust, per person or per product. Every step is one click, and the stop button halts everything.
- 1
Shadow
Nothing changes. Every action is shown as a preview.
- 2
Manual
Every action that changes anything waits for you.
- 3
Plan approval
You approve each day's plan once; the actions inside it then run.
- 4
Guarded auto
Routine work runs. High-risk actions, like merging to production, wait for you.
- 5
Auto with veto window
High-risk actions wait 15 minutes, during which you can cancel them.
- 6
Full auto
Everything runs. Budgets and policies still apply.
Built in
- Five checks before any action leaves
- Daily and monthly spending limits
- Each organisation isolated in the database
- A full record of what happened and why
- One button that stops the whole team
Give your goal a team.
Questions
Who actually does the work?
AI employees, each with a role: a Product Owner, a Marketer, a Developer and more. They use the tools you connect and act only through the same checked path as everyone else.
Can it do something I did not approve?
Only if you allowed it. Six autonomy levels run from “preview only” to “everything runs”. Spending limits, rules and the stop button apply at every level, and risky actions are checked before they happen.
Which models does it use?
A small fast model does the work and a second model checks it. You do not pick or pay for them separately during early access.
Can people work in it too?
Yes. People, AI employees and bots share channels, meetings, the inbox and the office. A person can take any task, and any task can be handed to the team.
Is my data separate from other teams?
Every organisation is isolated in the database itself, not only in the application. Secrets are stored encrypted and never shown to the team’s models.
How do I get in?
Growth HQ is invitation-only in early access. Create an account with the email you were invited on.