Rate limiting is a cross-cutting concern, not a product. It shows up wherever traffic has to be shaped, and in an agentic world that is everywhere. When an autonomous client can fire requests as fast as the network allows, the rate limit stops being a footnote in the docs and becomes the contract that keeps the whole system standing. The Rate Limiting tag on APIs.io gathers the providers and APIs where that concern is explicit.
Who carries the tag
The tag spans 13 providers and 6 APIs on the network. The providers cluster around gateways and traffic control, which is exactly where you would expect rate limiting to live: Arcjet, Cloudflare AI Gateway, Google Cloud Endpoints, Kong AI Gateway, Lunar, Spring Cloud Gateway, Stellate, Unkey, and RapidAPI all appear, alongside Descript, Public Sector, and the University of California, Berkeley.
The APIs tagged make the concern concrete. Kong AI Gateway publishes an AI Rate Limiting Advanced Plugin. Lunar carries its API Consumption Gateway. RapidAPI carries its Gateway API. UC Berkeley exposes a University API Gateway, and Public Sector contributes api.data.gov, the shared key and rate-limit layer behind a swath of US federal APIs. Descript rounds out the set with its Platform API.
The tag’s nearest neighbors in the catalog are API Management and API Gateway, which confirms the pattern. Rate limiting is not a standalone capability so much as a facet of managing and gating an API surface.
One observation
Here is the gap worth naming. Almost everything under this tag documents rate limits in prose, a plan tier, a requests-per-minute figure buried in a pricing table. Far fewer expose those limits where an agent can read them: machine-readable headers, a limits endpoint, or a policy artifact the client can inspect before it starts firing. A documented limit tells a human what to expect. A machine-readable limit tells an autonomous client when to back off without being told. As agents become the dominant consumers, that difference decides whether your integration degrades gracefully or hammers into a wall of 429s.
Takeaway
Rate limiting is thin as a standalone tag but foundational as a discipline, and the providers carrying it are overwhelmingly the gateways that enforce it for everyone else. Browse the Rate Limiting tag to see the full set, and when you evaluate any of them, ask not just what the limit is but whether your code can discover it at runtime.