Cost management

Calls to metered third-party APIs add up fast. HARP sits as a forward proxy between your applications and the remote APIs they call, so you can cut the volume of billable requests without changing application code.

Understanding the use case

Many widely used APIs bill per request or per processed unit: OpenAI for model calls, Google Maps for geocoding, Amazon Polly, Cloud Vision, Comprehend, Rekognition, or DeepL, among others. When several services call the same endpoints with similar inputs, the same paid request is often made again and again. Because HARP sits in front of those APIs, it can serve a previously fetched response instead of forwarding a new billable call.

Reducing billable calls with caching

HARP ships an RFC-9111 compliant HTTP caching system. When an API marks a response as cacheable, HARP stores it and serves it directly on subsequent matching requests, reducing remote traffic and the number of paid calls that reach the backend.

Because the cache lives in the proxy rather than in each application, a single cached response is shared across every consumer behind HARP. Repeated requests for the same data, common in microservice architectures, collapse onto one upstream call.

Many real-world APIs are not cache-optimized and omit usable cache headers. For those, the HARP rules engine can override response headers (for example, setting Cache-Control) so results can be cached according to your own freshness requirements.

Controlling traffic with the rules engine

The HARP rules engine runs small scripts at each stage of the request lifecycle. Beyond adjusting cache behaviour, it can short-circuit a request and return an immediate response without forwarding it to the remote API. That lets you reject or answer locally the calls you do not want to pay for, before they ever reach a metered endpoint.

Visibility

HARP exposes its traffic through a dashboard and through observability metrics. Reviewing transactions flowing through the proxy helps you see which endpoints are called most often and where caching or short-circuit rules will have the most impact on cost.

Benefits

• Fewer billable calls: caching and short-circuit rules keep repeated and unnecessary requests from reaching metered APIs.
• Shared results across consumers: one cached response in the proxy serves every application behind it.
• No application changes: HARP applies these controls at the proxy layer, so callers stay untouched.

Conclusion

HARP helps control the cost of metered API usage by caching cacheable responses, letting you cache APIs that are not cache-optimized through the rules engine, and short-circuiting requests you do not want to forward. All of this happens in the proxy, without modifying the applications that make the calls.