Pricing is where most sellers leave money on the table — or scare buyers off. An AI agent isn't a prompt with a bigger number on it; it's a different product with a different cost structure and a different promise. This guide gives you a practical way to price agents, assistants, and the heavier building blocks so the number feels fair to the buyer and worth it to you.
§ 01 First, separate the two costs
Buyers conflate two things. You shouldn't.
Your price is for the agent — the design, the tested runbook, the taste baked into it. The running cost is what the buyer pays the model provider in tokens every time the agent works.
Agents are the heaviest building block: more steps, more tokens, more time. A buyer who understands that the running cost is theirs (and ongoing) will judge your price more fairly. Point them to How we count tokens so the total cost of ownership is clear up front — surprise costs are the fastest route to a refund request.
§ 02 Price the outcome, not the effort
It's tempting to price by how long the agent took you to build. Buyers don't care how hard it was; they care what it does for them. Anchor on the outcome:
- What does one successful run save them? An agent that drafts a week of social posts saves hours of someone's time. Price against that, not against your build hours.
- How often will they run it? A one-off audit agent and a daily-use research agent justify very different prices.
- What's the downside if it's wrong? Agents that touch money, customers, or published output carry more responsibility — and more value when they're reliable.
A useful gut check: if the agent saves a professional two hours a month, almost any one-time price under that hourly rate is an easy yes.
§ 03 Three ways to land on a number
Use whichever fits — most sellers blend them.
- Value-based. Estimate the buyer's saved time or money per month, then price at a fraction of it. Highest ceiling, needs the most confidence.
- Comparable-based. Look at similar listings on the marketplace and position deliberately. On a curated platform, undercutting signals lower quality as often as it wins the sale.
- Cost-plus floor. Whatever you choose, it should clear your own bar for the work — a price so low it isn't worth supporting will leave you resentful and the listing neglected.
Set the value-based number as your target and the cost-plus number as your floor. The right price is usually between them.
§ 04 What buyers actually pay for in an agent
When a buyer compares two agents, the higher-priced one usually wins on proof, not features:
- A tested runbook. Sigrix listings show the agent's tested steps so buyers see exactly what it'll do. Visible, repeatable behaviour justifies a premium.
- Clear boundaries. "Here's what it does, here's what it won't touch" reduces the buyer's perceived risk — and risk is what they're really pricing.
- Model and version support. Knowing it's tested on the model they use removes a reason to hesitate.
- Reviews and a trustworthy seller. Social proof and the curated badge let you charge what an unvetted listing can't.
Every one of these is a lever you can pull to move the price up without changing the agent itself.
§ 05 Tiering and bundling
You rarely need one price — you need a ladder:
- Single block: a standalone agent or assistant at a focused price.
- Bundle (a product): package a persona, a few skills, and the agent into one listing with one price. Bundles raise the average sale and are easier for buyers to choose than assembling parts themselves.
- Good / better: a lean version and a fuller version let buyers self-select. Many will pick the more expensive one simply because it exists.
Bundling is also how a prompt seller grows into an agent seller — start with the prompt playbook, then compose upward.
§ 06 Adjust with real data
Your first price is a hypothesis. After launch:
- Watch conversion, not just views. Lots of views and no sales usually means a trust or clarity gap, not a price that's too high.
- Read the reviews. "Worth it" and "wish it did X" both tell you whether there's room to raise the price or add a tier.
- Raise deliberately. If it's selling well and reviewing well, a measured increase is normal — you priced the early version, not the proven one.
And keep the economics on your side: you keep 80–85% of every sale, with founding sellers locked at the lowest fee. Price for the value you deliver, be transparent about the running cost, and let the proof on the page carry the number.
FAQ
How do I price an AI agent? Anchor on the outcome — what one successful run saves the buyer — then position against comparable listings on the marketplace and set a floor that clears your own bar for the work. Price the value of the job done, not your build hours; the running token cost is the buyer's and stays separate from your price.
What is the difference between an AI agent's price and its running cost? Your price is a one-time charge for the agent itself — its design and tested runbook. The running cost is what the buyer pays the model provider in tokens each time the agent runs. Keep the two separate and explain both so there are no billing surprises.
How much should I charge for an AI agent? More than a prompt, because agents do multi-step work. Start from the value of one successful run, position against comparable agents, and never price below the point where supporting the listing is worth your time. Raise deliberately as reviews come in.
Should an AI agent cost more than a prompt or assistant? Usually yes. An agent runs autonomously across multiple steps and tokens, so it does the most work and carries the most responsibility — exactly what buyers pay a premium for. Price for the outcome and the proof on the page, not the listing type alone.
How do I raise the price of an AI agent later? Watch conversion and reviews, not just views. If it sells and reviews well, a measured increase is normal — you priced the early version, not the proven one. Adding a "good / better" tier lets buyers self-select into a higher price without repricing the original.
Keep reading
- How to sell AI prompts (and get paid for them) — the starting-point playbook.
- The five Sigrix building blocks — where agents fit.
- How we count tokens — explain running cost to buyers.
- Ready to list? Start selling on Sigrix.
One practical essay, every other Tuesday.
Patterns, sample listings, and short essays from the Sigrix editorial team. One-click unsubscribe, no noise.