How your team of agents works
Specialist agents do real work for you, and anything risky waits for your review before it goes out.
Your agents are not one chatbot. They are a small team of specialists, each with a role, tools, and limits. They do real work — writing, researching, coding, triaging — and they hand it back to you with evidence.
A team, not one bot
A typical setup gives you a few specialists. For example:
- An assistant that keeps your day organised, holds your personal context, and protects your attention.
- An engineer that writes and reviews code, checks for risk, and runs the tests.
- A company operator that handles the shared inbox, triages customers, and keeps the team in the loop.
You can add, remove, or rename roles to match how your company actually works.
How they handle bigger jobs
When a job is large or touches something sensitive, your agents do not just start changing things.
- They prepare the work first instead of guessing.
- They run it in a safe, separate copy so your live setup stays untouched.
- They show you the result and the proof before it counts as done.
Review before anything risky
Nothing that leaves your company or changes something important goes out on its own. Sending an outside message, publishing, changing data, or spending money pauses for your approval. Low-risk work — reading, drafting, organising — can go ahead inside the limits you set.
Proof, not promises
A job is not finished because an agent says it is. It is finished when you can see the result, compare it with what you asked for, and decide it is good enough.