// the find
kubeshop/testkube
☸️ The Open Testing Platform for AI-Driven Engineering Teams
Testkube is a Kubernetes-native test orchestration platform that runs any test framework (k6, Playwright, Cypress, etc.) as a workflow inside your cluster. The open-source agent is MIT-licensed and runs standalone; the control plane with multi-cluster visibility and the AI features are commercial. Aimed at platform engineering teams who want test execution to live in the cluster rather than in CI runners.
The TestWorkflow CRD is a genuinely good abstraction — it gives you Kubernetes-native test scheduling without forcing you to rewrite tests; existing scripts just work. The trigger system (k8s events, schedules, webhooks, MCP) is well thought out and covers most CI/CD integration patterns without requiring a pipeline rewrite. The OSS agent runs completely standalone with no phone-home requirement, which is rare in this category — most competitors gate that behind a login. The MCP server addition means you can wire test execution into an AI agent loop without building glue code.
The open-source/commercial split is aggressive: multi-cluster visibility, AI analysis, and RBAC are all paywalled, so the OSS tier is essentially a single-cluster trial with no upgrade path short of signing up for SaaS. The repo shows a heavy v1/v2/v3 API version history for core types (Tests, TestSuites, TestWorkflows) — that's a lot of migration surface and signals the data model has changed substantially; adopters who built on v1 executors have real upgrade work ahead. Star count (1.6k) is surprisingly low for a project with this much enterprise backing and named customers like Nvidia and T-Mobile, which suggests organic community traction is weaker than the company narrative implies. Local development requires Tilt and a running Kubernetes cluster, which is a non-trivial barrier for contributors compared to projects that can spin up with docker-compose.