From a sentence to a working team — in five steps.
You describe a goal in plain language; Bro assembles a small team of specialized AI agents and shows you the design before anything runs. Agents find, judge and draft on a visible canvas, and every outward action waits for your one-tap approval. As trust builds, you can grant routine autonomy — and take it back anytime.
Your part
plain language, not templates
Describe your goal
In plain language — "20 new cleaning clients in Amsterdam-Zuid" beats "more customers". Bro asks one to three short questions (audience, region, tone, what success looks like) so the team fits you, not a template.
“Get more clients for my cleaning business in Rotterdam”
Finder
Finds local businesses that fit
Qualifier
Scores each site, keeps the best
Outreacher
Drafts a personal opener
Bro's proposal
approved before anyone starts
Meet your team
Bro proposes a small team — usually two to seven agents, each with one sharp responsibility. A lead-gen goal gets a Finder, a Qualifier and an Outreacher; a reviews goal gets a Spotter and a Responder. You approve the design before anyone starts.
On the canvas
watch it think
Agents do real work
The Finder gathers matching businesses, the Qualifier reads each website and scores it 0–100 with a written reason, the Outreacher drafts a first message that opens with a real finding. Every agent's reasoning is visible live on the canvas — you can watch it think.
Qualifier
Reads each site, keeps the best
Qualifying
Checking mobile, HTTPS, load speed and freshness…
The gate
one queue, server-enforced
You approve the important moves
Every outward action — an email, an SMS, a post, spend — lands in one Approvals queue with the full context and the agent's why. Approve with one tap, edit first (the agent learns from your edit), or dismiss. This gate is server-enforced, not a UI promise.
Outreacher drafted an email
To: Sparkle Clean · sparkle-clean.nl
Hi — I noticed your booking page takes ~6s to load on mobile, which is where most of your customers are. I put together a couple of quick wins…
Why: opens on the site's weakest point, in your tone.
The long run
step back down anytime
Earn autonomy on your terms
Start with approve-everything. When an agent has earned it, grant routine autonomy per ecosystem — independent within limits, 24/7, with a daily budget, send window and do-not-contact list still enforced. Step back down anytime; there's also a kill-switch.
12
Drafts waiting
34
Approved & sent
14
Reviews answered
6h 20m
Time saved
The autonomy ladder
per ecosystem, reversible
Autonomy is earned, never assumed.
Rung 1
Approve everything
Every outward action queues for you. This is where every ecosystem starts — you see exactly how the agents judge and write before trusting them with more.
Rung 2
Routine autonomy
Grant independence for the repetitive, low-risk work — per ecosystem, within hard limits: a daily budget, a send window and the do-not-contact list stay enforced.
Rung 3
Autonomous within limits
Agents run 24/7 on your terms. Everything remains on the timeline and in the audit log, and one switch pauses all autonomous work instantly.
You approve everything
Agent sends the routine ones
Fully autonomous, on your terms
Good to know.
How long does setup take?
Minutes, not days. You describe the goal, answer a couple of clarifying questions, and approve the proposed team. Finding leads works out of the box; sending email requires connecting your own sender (Resend or Gmail), which is a one-time step.
What exactly requires my approval?
Everything outward-facing or irreversible: sending an email or SMS, publishing a post, spending budget. Internal work — finding, scoring, qualifying, drafting — runs on its own, because it's reversible and visible on the timeline.
What does "server-enforced approval" mean?
The send action reads the recipient and content from the database and passes guardrails (daily limit, send window, do-not-contact list) on the server. A bug or a prompt trick in the interface can't skip the gate, and every approved action is written to an audit log.
Can agents work while I'm offline?
Yes. Running agents do one small unit of real work at a time, server-side, around the clock. Anything outward still queues for your approval — so you come back to drafts and findings, never to surprises.
What if an agent drafts something off-brand?
Edit the draft before approving; the agent distills your edit into a lasting lesson and applies it to future drafts. You can also teach agents rules directly ("never open with flattery"), which they follow from then on.
Do I need my own AI API key?
No. Bro and the agents are built in, powered by Claude. The only keys you ever connect are your own channels — like your email sender — stored in an encrypted vault and read server-side only for actions you approved.
See the five steps run on your goal.
Free 7-day trial with 500 credits. No credit card. Nothing goes out without your approval.