API Gateway Pricing: How the Models Actually Work
API gateway pricing usually falls into three buckets: per-request/per-call pricing, per-seat or per-user pricing, and flat-tier pricing with usage caps. Most vendors mix two of these — a base subscription plus overage fees, or a seat price plus a request allowance — which makes comparing vendors harder than it should be. The real cost isn't the sticker price on the pricing page; it's what happens when your traffic or team size doubles.
This article breaks down how each model works, what actually drives your bill up, and what to check before picking a gateway — whether you're routing REST traffic between microservices or exposing an AI model as an API.
The Three Common Pricing Models
1. Per-request (metered) pricing
You pay per million API calls, sometimes tiered so the marginal cost drops at volume. This is common for cloud-native gateways like AWS API Gateway or Azure API Management's consumption tier. It scales naturally with usage but makes costs unpredictable — a traffic spike, a retry storm, or a misbehaving client can turn into a surprise invoice at the end of the month.
2. Per-seat pricing
You pay per developer, per team member, or per "active user" who has access to the dashboard or the keys. This is common for API management platforms aimed at teams, and for SaaS products that wrap an underlying API (like turning a shared account into individual API keys per person). It's predictable and easy to budget, but it can penalize small teams with heavy usage, or overcharge large teams with light usage.
3. Flat-tier pricing with included usage
You pay a fixed monthly fee for a tier that includes a request allowance, a number of seats, or both, with overage charged separately once you exceed it. This is the model most SaaS API gateways use because it balances predictability with fairness — light users pay less, heavy users pay more, but nobody gets a $4,000 surprise bill from a traffic spike they didn't cause.
What Actually Drives Cost
Before comparing sticker prices, look at what actually contributes to your bill:
- Request volume — raw call count, sometimes billed per million.
- Data transfer — some gateways charge separately for egress bandwidth, which matters a lot if you're streaming large responses.
- Seats/users — how many people need their own API key or dashboard access.
- Features gated by tier — rate limiting, custom domains, streaming, tool/function calling, and usage analytics are often locked behind higher plans.
- Support tier — SLA and support response time are frequently the real differentiator between a €9/month plan and a €49/month plan, not the underlying infrastructure.
If you're comparing two gateways with similar per-request costs, the deciding factor is usually which features are included at your expected usage level, not the headline number.
A Practical Cost Comparison
Say you have a five-person team, each needing their own API key, making a combined ~2 million requests a month with occasional streaming responses and tool calls.
Per-request model: at $1–$3.50 per million requests (typical for cloud gateways), that's $2–$7/month in pure compute cost — but you still need to build and maintain your own key management, per-user isolation, rate limiting, and streaming support yourself. The "cheap" number hides the engineering time.
Per-seat model: five seats at, say, $15–$20/seat/month puts you at $75–$100/month regardless of whether you use 200,000 or 2 million requests. Predictable, but you're paying for headroom you might not need.
Flat-tier SaaS model: a tiered plan with per-seat pricing that already includes streaming, tool use, and usage metadata avoids the build-it-yourself tax entirely. For example, SubToAPI's pricing runs Solo at €9 for a single key, Team at €19/seat, and Scale at €49/seat for larger teams needing higher limits and priority support — with a free trial at signup so you can measure your actual usage before committing.
The point isn't that any one model is objectively cheaper. It's that the right model depends on whether your usage is spiky or steady, and whether "seats" or "requests" better reflects how your team actually consumes the API.
Questions to Ask Before You Commit
- Is streaming billed differently than standard requests? Some gateways treat a long streaming response as one request; others meter it by duration or token count. This can 5–10x your effective cost if you don't check.
- What happens on overage? Hard cutoff, automatic throttling, or pay-as-you-go overage billing — these have very different failure modes in production.
- Are seats per-person or per-key? If a seat corresponds to one shared key used by five scripts, you're paying differently than if each script needs its own key.
- Does the plan include tool/function calling and usage metadata, or is that a separate add-on? If you're building an agent, this matters more than raw request price. See /docs/tools for how tool calling is typically exposed via a gateway API.
Getting Started Without Overcommitting
If you're evaluating a gateway to sit in front of an existing AI subscription — turning a Claude account into a proper HTTPS API with its own keys, streaming, and usage tracking per team member — start with a free trial rather than estimating from a pricing page. Actual usage patterns (request size, streaming frequency, team size) matter more than any published tier. SubToAPI's quickstart gets you an API key issued in a few minutes so you can benchmark real traffic against real pricing tiers before choosing a plan.
FAQ
Is per-request pricing always cheaper than per-seat pricing? Not necessarily. Per-request pricing looks cheaper on paper for low volume, but it doesn't include the engineering cost of building key management, rate limiting, and monitoring yourself. Per-seat or tiered SaaS pricing usually bundles those features in.
What's a normal price range for API gateway plans? Individual/solo plans typically run $5–$15/month, team plans $15–$30 per seat/month, and enterprise-grade plans with SLAs and dedicated support start around $50+ per seat/month or scale into custom pricing for high volume.
Does API gateway pricing include streaming and tool calls? It depends on the vendor — some charge extra or gate these behind higher tiers. Always confirm streaming (/docs/streaming) and tool use (/docs/tools) are included in the plan you're evaluating, not sold as add-ons.