What an Agent Fleet Actually Costs a Pre-Revenue Founder

Most of the content about AI agents talks about what they can do. I want to talk about what they cost, because if you are a pre-revenue founder looking at this space, the cost question is the one that actually matters to your bank account.

Armada Works is an agent-first consultancy. We deploy fleets of Claude Code agents into client codebases, run them, and hand the system over. I run the same kind of fleet against my own business, so I know what it costs from both sides of the engagement: as the operator building the system, and as the founder paying the cloud bill. This post is the honest accounting. No case studies, no ROI claims, no "10x your output" language. Just what you spend, what you get, and whether it is rational at this stage.

The Engagement Fee: What You Pay Armada

Start with the direct cost. Armada's entry point is a Pilot engagement: one agent, shipped into your repo over five working days, for $2,500 to $4,000. The Pilot is designed to answer a specific question: does an agent-based approach solve your particular bottleneck? The scope is narrow on purpose. You get a single agent running a defined task (content, outbound, SEO, whatever your constraint is), a runbook documenting how it works, and a handoff document explaining how to keep it running. If you continue into a longer engagement within 30 days, the Pilot fee credits 100% toward that engagement.

The longer engagements come in three shapes. The Operate tier ($5,000 to $12,000 per month) means we run a four-to-six agent fleet for you on an ongoing basis. The Build tier ($15,000 to $60,000 per project, or $8,000 and up per month as a retainer) is for agent-assisted product engineering, not just marketing. The Transfer tier ($10,000 to $20,000 one-time, with optional ongoing support at $1,500 per month) is the one I think matters most at the pre-revenue stage: we build the fleet, we hand it over, and you run it yourself. The explicit goal is that you stop paying us.

Those are the Armada-specific numbers. They are published on the pricing page, and they will not surprise you on a call. But they are not the full picture.

The Infrastructure Layer: What You Pay Everyone Else

An agent fleet runs on the same cloud infrastructure your product does. The agents themselves are Claude Code sessions running against the Anthropic API, so your API bill scales with how often the agents run and how much context they process. Beyond that, you need the same things any modern web application needs: a database, a hosting platform, a domain.

I am not going to quote specific infrastructure numbers here because they depend on your stack, your scale, and how aggressively you configure caching and model selection. What I can tell you is that the infrastructure cost is small relative to the engagement fee. If you are running a fleet of five to seven agents on a reasonable schedule (three runs per week per agent, not continuous), your API and hosting costs are a fraction of what you pay the consultancy. The cloud bill is not the thing that should give you pause. The engagement fee is the real number, and I want you to evaluate it on its own terms.

What the Fleet Actually Produces

Here is where most vendors start talking about ROI, and here is where I am going to refuse. Armada has zero paid clients. I am not going to tell you that our agents generate a certain number of leads per week or increase your traffic by some percentage, because I do not have that data from a paid engagement. What I can tell you is what the fleet produces as output, because I see it every day against my own business.

A five-agent marketing fleet running three times per week produces roughly this: two to three blog posts per week, each 1,200 to 2,500 words with SEO optimization. One keyword audit per week with concrete recommendations. Outbound prospect research for five to ten leads per week. Inbound lead triage within the same business day a lead comes in. A daily synthesis from the CMO agent that tells you the three to five things worth your attention.

That is the throughput of the system. Whether it translates to revenue depends on your market, your positioning, your offer, and a dozen other things that no agent fleet controls. The fleet handles execution. Strategy still sits with you.

The Honest Math for a Pre-Revenue Founder

Here is the calculation I think matters. If you are pre-revenue, you are trading money for time. Every dollar you spend is a bet that the thing it buys will either generate revenue faster or save you time you redirect toward generating revenue. An agent fleet is a bet on the second kind. It does not bring in money directly. It reclaims the hours you are currently spending on content, SEO, outbound, lead triage, and operational coordination.

For a solo founder, those hours are real. I was spending fifteen to twenty hours a week on marketing operations before the fleet was running. Writing blog posts. Researching keywords. Sending cold emails. Reading every dashboard. That time now goes into product work and conversations with prospects. The fleet costs money, but so does my time, and the time cost was not showing up on any invoice.

The question is not "will this pay for itself in month one." It will not. The question is whether reclaiming those fifteen to twenty hours per week, applied to product and revenue-generating work, compounds into enough value over three to six months to justify the spend. That is a judgment call, not a spreadsheet calculation. It depends on what you do with the hours.

When It Is Not Rational

There are cases where this math does not work, and I would rather tell you now than discover it on a call.

If you do not have a marketing bottleneck yet, you do not need a marketing fleet. If your constraint is product-market fit, spending $5,000 to $12,000 per month on marketing operations is premature. Get the product right first.

If your team will not read diffs in main, the entire coordination model breaks. The fleet commits its work to your repository. If nobody reviews the commits, nobody catches drift, and the agents will happily produce output that wanders from your positioning. An agent fleet requires a human in the loop who looks at the work regularly. Not hourly, but at least weekly.

If you want a contractor who does what you say, an agent fleet is the wrong framing. Agents run on prompts and constraints, not live direction. You set the system up, you review the output, and you adjust the prompts when the output drifts. This is closer to managing a team than managing a freelancer. If you want someone to write a blog post on Tuesday and send an email on Thursday, you want a VA, not a fleet.

What I Would Do in Your Position

If I were a pre-revenue founder evaluating this, I would start with the Pilot. $2,500 to $4,000 for one week, one agent, one concrete answer about whether this approach fits your bottleneck. You keep the agent, the runbook, and the handoff doc regardless of whether you continue. The Pilot fee credits toward the longer engagement if you do. There is no follow-up sequence, no drip campaign, no sales pressure after the call. We talk for thirty minutes, decide together whether agents are the right fit, and either proceed or part ways.

That is the honest version. If this sounds like the kind of economics you want to discuss in concrete terms, book a discovery call and we will walk through the numbers against your specific situation.