Kubernetes-native execution engine
Fast, reusable execution engine on Kubernetes
kruntimes helps platform teams and developers run functions, pipelines, batch jobs including AI workloads, and agent sandboxes on warm Runtime Pod pools without paying a Pod cold start for every execution.
kruntimes is a Kubernetes-native execution engine for serverless functions, CI pipelines, batch workloads including AI workloads, and AI agent sandboxes. It keeps Runtime Pods warm and assigns fine-grained Runs to those pools without creating a new Pod for every execution.
The project is currently v0.x experimental with v1alpha1 APIs. Built-in
Runtimes are intended for trusted workloads in trusted namespaces.
Start Here
- Project overview
- Quick Start
- Installation
- Usage guide
- Architecture
- API reference
- Configuration
- Troubleshooting
- FAQ
- Development guide
- Testing guide
- Operations guide
- Security and threat model
- Custom Runtime development
- Compatibility matrix
- Release process
- Performance benchmarks
- Roadmap
- Community and governance
Understand the project
Problem, approach, value, use cases, and current limits.
Run the first workload
Start a local environment and execute a Bash Run.
Operate kruntimes
Install, upgrade, uninstall, troubleshoot, and recover a cluster deployment.
Review the security model
Understand trusted workloads, namespace boundaries, and current isolation limits.
Build a Runtime
Implement the local Runtime Server API and execution lifecycle.
Benchmark capacity
Measure scheduler latency, throughput, Runtime slots, and API load.
Contribute
Set up a development environment, run tests, and open PRs.
Join the project
Find support, security reporting, governance, and conduct docs.