Speedscale

Speedscale captures real production traffic and replays it to validate code before it merges, positioned around validating AI-generated code, with an eBPF collector for debugging encrypted microservice traffic and observability from capture through to validated fixes.

Generating simulations from captured traffic rather than writing them is the fundamental difference from every hand-authored mocking tool, and it addresses the reason simulations go stale. A mock written in March reflects the API as it was in March, while a capture from last Tuesday reflects what the service actually does now, including the malformed requests nobody would have thought to write.

Positioning around AI-generated code is a current and defensible framing rather than opportunism, because the volume of code reaching review has risen faster than the capacity to review it, and replaying real traffic against a change is one of the few validation methods that scales with generated volume.

The eBPF collector is the technically notable piece, since inspecting encrypted service-to-service traffic normally means terminating TLS or instrumenting every application, and observing it in the kernel avoids both. Rates are not published on the pricing page. Speedscale runs alongside services in Kubernetes rather than as a separate environment to maintain.