What to charge for your
MCP server tools.
Per-call pricing is unfamiliar territory for most builders. This guide gives you category benchmarks with the reasoning behind them, a framework for setting the number, how agent buyers actually behave on price, and a worked revenue projection — so you can price with confidence instead of guessing.
Pricing an MCP tool is different from pricing a SaaS plan. There's no seat, no tier, no monthly minimum — just a price per call, paid in USDC, settled the instant the tool returns a result. That simplicity is the whole point, but it means the number itself carries all the weight.
The good news is that per-call pricing is forgiving. The floor is $0.01, the price is one field, and the market tells you within hours whether you got it right. You don't have to nail it on the first try — you set a sensible starting point and let real call volume tune it.
This guide covers four things: what comparable tools charge, how to reason about your number, how price-sensitive agent buyers are, and how to project revenue.
Benchmarks by category
These are typical per-call ranges for x402-ready tools today. Treat them as a starting point, not a ceiling — a tool that saves an agent a slow, expensive workaround can charge well above the band.
Search and lookup ($0.01–$0.03): commodity capability, lots of competition, so price near the floor and win on volume. Data and enrichment APIs ($0.01–$0.05): value scales with freshness and coverage. Document extraction and OCR ($0.02–$0.10): real compute per call justifies the higher floor. AI inference and generation ($0.02–$0.25): price tracks model size and tokens returned. Compliance checks like KYC/AML ($0.05–$0.50 per check): high value, low volume, low price sensitivity. Specialized or financial data ($0.05–$1.00): the more proprietary the source, the more pricing power you have.
Anchor on value, not cost
The instinct is to price off what a call costs you to serve. Resist it. Anchor instead on the value of one successful call to the agent. If your tool replaces an hour of an agent's fumbling, or substitutes for an API the agent would otherwise pay $0.50 to hit, then $0.05 is a bargain that still earns you a healthy margin.
If the call is a commodity — a lookup the agent could get five other places — the value anchor is low and you should price near the floor and compete on reliability and latency instead.
How agent buyers behave on price
Your buyer is an autonomous agent on a budget, not a human with a credit card and brand loyalty. Agents are price-sensitive in aggregate and will often pick the cheapest option that clears a quality bar. A tool priced 10x its peers gets called a fraction as often, regardless of marketing.
That makes pricing slightly below the obvious alternative a reliable volume-maximizing move — especially early, when you're building a base of agents that habitually route to your tool. Latency and success rate matter too: an agent that gets a fast, correct result will keep coming back and will tolerate a modest premium for it.
Projecting revenue
Revenue is simply price × successful calls, minus Mailgent's per-call fee. The two obvious levers trade off against each other: at $0.01 a call, 100,000 calls a month is $1,000 of gross; at $0.05, you only need 20,000 calls to reach the same $1,000. Your job is to find the point on that curve where price and volume multiply to the most.
Don't model this blind. The seller pricing simulator lets you plug in a price and a monthly call volume and shows exactly what you keep after Mailgent's fee and credit costs — so you can compare scenarios before you commit to a number.
Start near the floor, then raise
If you're unsure, list at or near $0.01. A low entry price is the cheapest customer-acquisition tool you have: it pulls in first callers and produces the usage data you need to price properly. Once demand is clear and steady, raise the number toward the value you're delivering.
Because repricing is a one-field edit and fully reversible, there's no penalty for adjusting as you learn. Raise until volume tells you you've gone too far, then settle one step back.
Per-tool pricing
Pricing is per tool, not per server. Each requirePayment wrapper carries its own amount, so a cheap commodity lookup and an expensive generation endpoint can live on the same server at very different prices. Price each tool on its own value rather than averaging across the server.
FAQ
What currency do I get paid in?
USDC on Base. Settlement is instant per call, and you can cash out to any wallet you control.
Is there a minimum price?
Yes — $0.01 per call. There's no maximum; price to the value your tool delivers.
Can I price different tools differently?
Yes. Pricing is per tool — each requirePayment wrapper has its own amount — so a cheap lookup and an expensive generation can sit on the same server at different prices.
What if I price too high?
Call volume drops and you'll see it within hours. Lower the number — it's a one-field change — and volume recovers. Per-call pricing is fully reversible, so overshooting costs you a little revenue, not a relationship.
How does Mailgent's fee affect what I keep?
Mailgent takes a per-call fee on paid calls; you keep the rest. The simulator on the seller pricing page shows your take-home at any price and volume so you can price with the fee already accounted for.
Should I bundle or charge per call?
Per call is the native model — agents pay for exactly what they use, which is what makes them willing to try your tool with no commitment. If you want volume incentives, the cleaner lever is simply a lower per-call price.
Related reading
More mailgent index
Last updated: 2026-05-27