Speedscale
API traffic capture and replay with AI bug-repair agents.
About
Speedscale captures live HTTP, gRPC, and SQL traffic via eBPF inside your Kubernetes cluster — language-agnostic, no code changes — then replays it against new service versions to validate functional behavior and performance SLOs in one run. It auto-generates service mocks from recorded traffic (REST, gRPC, Postgres, Kafka), and its newest angle is AI bug repair: capture, triage, reproduction, and validation agents that run entirely in your cloud so production payloads never leave your perimeter. A free proxymock CLI gives AI coding agents real production context via MCP.
It is a commercial SaaS product strongest for Kubernetes teams, and the autonomous repair loop is new enough that human-gated mode is the sensible default.
What it can do
- eBPF traffic capture from production or staging
- Traffic replay with latency, throughput, and error-rate metrics
- Snapshot testing against recorded baselines
- Auto-generated service mocks from traffic
- AI bug-repair agent loop (capture, triage, reproduce, validate)
- Free proxymock CLI with MCP for coding agents
Setup
- Install the Speedscale sidecar or node agent in your Kubernetes cluster.
- Capture production or staging traffic via eBPF.
- Replay traffic against new service versions; compare snapshots against baselines.
- Optionally enable the AI bug-repair agents (human-gated or autonomous).
Prerequisites
A Kubernetes cluster; services speaking HTTP, gRPC, or SQL.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: eBPF capture/replay, snapshot testing, auto-generated mocks confirmed across independent comparisons and own docs; AI bug-repair agent loop and free proxymock CLI with MCP confirmed in recent coverage; Customer names (Home Depot, IHG, etc.) are vendor-reported; Commercial SaaS, strongest for Kubernetes teams; repair loop is new; Capped at 8: newest AI features have thin independent verification. License: Proprietary SaaS; proxymock CLI free..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 8.0/10
- Openness · license + docs
- 5.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · eBPF capture/replay, snapshot testing, auto-generated mocks confirmed across independent comparisons and own docs
- · AI bug-repair agent loop and free proxymock CLI with MCP confirmed in recent coverage
- · Customer names (Home Depot, IHG, etc.) are vendor-reported
- · Commercial SaaS, strongest for Kubernetes teams; repair loop is new
- · Capped at 8: newest AI features have thin independent verification
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.