Glossary / Agents and memory
What is Agents team?
Definition
An agents team is a set of AI agents that each take a role in a project, for example a writer, an SEO agent or a LinkedIn agent, and work from the same project facts.
How it works
In Growth OX each piece of prepared work says which agent made it, such as "Writer prepared this". The agents read the same brain and leave decisions, notes and handoffs in Memory, so when one stops, another can continue. Agents can use a local model, search the web and use the computer through the browser and operating-system tools. Each step that reaches outside waits for your approval.
A team can be shared as a file. When you import one, the app shows what it would create before anything exists, and everything starts off: routines are paused, connections are empty and skills are not attached. Secret-looking text is removed from the exported file, and you should read the imported text as if a stranger wrote it.
Where it delivers value
A solo founder can split a launch week into drafting, search work and posting preparation, and see which agent prepared each piece, instead of one long chat where everything is mixed.
Limits
A team is not a staff replacement. Agents prepare; you approve anything that posts, sends or spends. A shared team file brings structure, not trust: connections and skills have to be set up and attached by you. Roles in a team are only as good as the project facts they read.
Compared with alternatives
| Option | Structure | Control |
|---|---|---|
| One chat assistant | A single thread | You steer every turn |
| Agents team | Named roles on one project | Approval rules on outside steps |
| Hiring help | People with their own judgment | Your management time |
Key takeaways
- Roles are visible: each piece says who prepared it.
- Agents share memory, so work can be handed over.
- Imported teams start with everything off.
FAQ
Can agents act without asking me?
Steps that reach outside wait for your approval.
What does a team file contain?
Agents, skills, paused routines and empty connection slots. Secret-looking text is removed.
Can another agent continue a stopped job?
Yes. Decisions and handoffs are kept in Memory, so another agent can continue from them.