How much does it cost to run AI agents? Credits, explained

Brohns has a single paid plan. Brohns Pro costs €39 per month and includes 4,000 credits; pay yearly and it's €390, which works out to two months free. Before you commit to anything, every account gets a 14-day free trial with 500 credits included — no credit card required. There are no seats to count, no per-agent fees, and no feature tiers to decode.

Credits are the meter. Every time an agent on your canvas actually does work — pulls in a batch of leads, judges whether one is worth pursuing, drafts an outreach email — a small, fixed number of credits moves. When your agents are idle, nothing is billed. The meter follows output rather than time, which suits freelancers and agencies whose workload swings from month to month.

One structural detail keeps the math simple: Claude, the model that powers every agent, is built into Brohns. You don't bring your own AI API key, and there's no separate usage bill arriving from a model provider. The credit price of an action is the entire price of that action.

There are five billable actions, and their prices never change mid-month. Everything an agent does maps onto one of these:

Notice the shape of the list: the expensive end is thinking work (sourcing, drafting), and the cheap end is execution. Sending an approved message costs a single credit because by that point the real work — research, judgment, writing — is already done. That's also why a qualification pass is only 2 credits: it's a focused yes-or-no judgment per lead, with a written reason, not a full research project.

Just as important is what's not on the list. Approvals, edits, briefs, and incoming replies cost zero credits. Credits only move when an agent actually does work — never when you review it, redirect it, or when a prospect writes back.

Scenario 1 — testing during the trial. Say you spend two weeks putting a lead-generation team through its paces: your Finder runs 4 sourcing batches (20 credits), the Qualifier does 4 qualification passes (8), the Outreacher drafts 40 messages (120) and you send all 40 after review (40), and 10 prospects reply, each getting a drafted response (30) that you approve and send (10). Total: 228 credits — less than half of the 500 the trial includes, with room left to experiment.

Scenario 2 — a freelancer working about 150 leads a month. The Finder runs 10 sourcing batches to surface roughly 150 leads (50 credits) and the Qualifier does 10 passes over them (20). You draft outreach to the 60 best-qualified leads (180) and send them after approval (60). Fifteen conversations come back; each reply that arrives is free, and each drafted response costs 3 (45) plus 1 to send (15). Total: roughly 370 credits — under a tenth of Pro's 4,000, which is why the fact that unused credits roll over one cycle actually matters.

Scenario 3 — a small agency running two client ecosystems. Thirty sourcing batches (150 credits), thirty qualification passes (60), 250 drafted messages (750) with 250 approved sends (250), and 50 active conversations with drafted replies (150) and sends (50). Total: about 1,410 credits — a genuinely busy month, still comfortably inside one Pro subscription. That's the honest version of "4,000 credits is enough to work hundreds of leads per month."

To estimate your own month, count actions, not hours: (sourcing batches × 5) + (qualification passes × 2) + (drafts × 3) + (reply drafts × 3) + (sends × 1). If you're not sure what your numbers look like yet, the trial's 500 credits exist precisely so you can find out with real work before paying anything. Writing a sharper goal up front also keeps agents from spending credits on the wrong leads — see how to write a goal AI agents can execute.

Brohns is built so that nothing leaves the building — no email, no post, no spend — until you've explicitly approved it, and that gate is enforced on the server, not just painted onto the interface. Pricing follows the same principle: the entire review layer is metered at zero. Opening the Approvals queue, reading an agent's reasoning, editing a draft, dismissing a bad idea, batch-approving twenty good ones — none of it touches your balance.

This isn't a giveaway; it's incentive design. If reviewing cost credits, careful users would pay a tax for staying in control, and the product would quietly reward rubber-stamping. Instead, the expensive thing is agent output and the free thing is your judgment — so you can be as picky as you like without watching the meter.

The free layer is also where quality compounds. When you edit a draft before approving it, the agent distills your change into a lasting lesson, and you can teach agents directly through briefs — all free. Over a few weeks, that means fewer edits, better drafts, and more of each credit going toward work you'll actually approve. If you later grant an ecosystem routine autonomy via the autonomy ladder, the per-action prices stay identical; autonomy changes who clicks send, not what the work costs.

Unused credits roll over one billing cycle. A quiet month — holidays, a big client project, a paused campaign — doesn't mean paying for capacity you didn't use; those credits are waiting when you ramp back up the next cycle. If you have an unusually heavy month instead, extra-credit top-up packs are the only add-on. There is no forced upgrade path and no second tier hiding the features you actually need.

A few cost-relevant facts about how the platform works: lead finding runs out of the box on the free OpenStreetMap path, with Google Maps as an optional upgrade. Sending email requires connecting your own sender — a Resend API key or Gmail via OAuth — because Brohns never sends from its own domain on your behalf. And your keys live in an encrypted credentials vault, so wiring up a sender is a one-time setup, not a recurring chore.

The bottom line: for €39 a month you get a team of specialized agents whose every action has a posted price, whose output waits for your sign-off, and whose meter only runs when real work happens. If you want to see what your first credits actually go toward day by day, start with your first week with an AI agent team — or open a trial account and check the numbers against your own pipeline, starting with lead generation.