kagent

kagent is a Kubernetes-native framework designed for building and managing AI agents within Kubernetes. It simplifies the process of deploying and orchestrating AI agents by leveraging Kubernetes custom resources for defining agents and tools. Agents consist of a system prompt, a set of tools, and a model configuration. Tools are external resources accessible to agents and can be shared across multiple agents.

Key principles include being Kubernetes-native, extensible, flexible, observable, declarative, and testable. The architecture comprises a Controller (Kubernetes controller managing custom resources), a UI for agent and tool management, an Engine (Python application built on Autogen for running agents and tools), and a CLI for command-line interaction.

The roadmap includes improvements in observability (tracing, oTEL integration, metrics), feedback/testing (eval framework, debugging, guided learning), runtime/engine enhancements (multi-agent support, graph execution, workflows, multiple LLM providers), and tool-related features (tool discovery, built-in tools as MCP server).