kagent

Why we included this project

Platform engineers and SRE teams who already run everything in Kubernetes will appreciate kagent as a way to treat AI agents like the rest of their workloads instead of throwaway scripts. It turns agents into declarative resources you define and version as CRDs, rolling them out through GitOps, and it manages the supporting pieces agents depend on: MCP tool servers, vector-backed long-term memory, and prompt templates. Because it supports multiple runtimes, including Go and Python ADK, and lets you bring your own agent framework like LangGraph or CrewAI, kagent keeps working as your stack changes. Teams use it to troubleshoot connectivity issues, debug Gateway configs, or automate Prometheus alert triage. Observability comes from OTel tracing and structured logs, and human-in-the-loop approval gates let you check before agents act on their own.