CNCF Sandbox framework for running AI agents natively in Kubernetes for automated diagnostics and cluster operations.
No compliance attestations on file. Confirm directly with the vendor before procurement.
| Tier | Price | Includes |
|---|---|---|
Open Source (Apache 2.0) | Free | Full feature set; community Discord support; CNCF Sandbox project |
Solo Enterprise for kagent | Contact sales | — |
kagent is a Kubernetes-native runtime for AI agents. It gives agents identity, policy, OpenTelemetry tracing, human-in-the-loop approval, lifecycle management, and a sandboxed deployment model for MCP tool servers — running on the same Kubernetes infrastructure platform teams already operate.
Platform engineering teams that need to take experimental agents (LangChain, CrewAI, Google ADK, A2A, custom Python) and run them in production with the same governance posture as their other workloads. Especially relevant for orgs already invested in Kubernetes, Istio, and CNCF tooling.
Deploy as a CNCF-style controller in existing clusters; bring agents from any framework and run them with unified identity and policy; deploy MCP tool servers as sandboxed Kubernetes resources via kmcp; trace agent activity end-to-end with OpenTelemetry; gate critical actions behind human approval; integrate with agentgateway for protocol routing (A2A, MCP) and agentregistry for skill discovery.
Early but well-architected. Backed by Solo.io's strong CNCF track record (Istio, Envoy, kgateway). Suitable for early production where you have platform-engineering ownership; less suitable as a turnkey hosted product.