What problem it solves

Building, deploying, and managing AI agents requires integrating agent behavior, runtime configuration, tools, models, sessions, and observability. The project addresses this with a declarative Kubernetes-based framework that supports multiple runtimes, model providers, and MCP tools.

Key features

• Declarative Agent Management — Agents are Kubernetes custom resources that combine behavior and runtime configuration. Configuration can be defined in YAML, kept in version control, and deployed through GitOps workflows.

  Source: README

• Flexible Runtime and Model Support — Agents can use the Go or Python ADK, Codex, Claude, or a custom runtime selected through a Harness. Supported LLM providers include OpenAI, Azure OpenAI, Anthropic, Google Vertex AI, Ollama, and custom providers and models accessible through AI gateways.

  Source: README

• MCP Tool Integration — Agents can connect to MCP servers that expose tools. The included MCP server provides tools for Kubernetes, Istio, Helm, Argo, Prometheus, Grafana, Cilium, and other systems, while RemoteMCPServer resources can be shared across agents.

  Source: README

• Persistent Sessions and Lifecycle Control — Sessions preserve conversations and task history across interactions, and agents can be suspended, resumed, checkpointed, and resumed from a checkpoint to fork a conversation. PostgreSQL stores sessions, tasks, and conversation history independently of running agents.

  Source: README

• OpenTelemetry Observability — OpenTelemetry tracing exposes agent and tool activity for monitoring. The framework supports tracing agent, model, and tool execution through an existing observability stack.

  Source: README

Who it may suit

This may suit developers and platform teams building AI agents that need Kubernetes-native configuration, deployment, and lifecycle management. It may also suit teams connecting agents to cloud-native tools, multiple model providers, or custom runtimes.

Topics to discuss

• Which agent runtime and model provider combinations should receive the most community testing and documentation?

• How should reusable AgentTemplate and RemoteMCPServer resources be organized for multi-team environments?

• Which observability workflows and troubleshooting signals should be prioritized for agent sessions and tool calls?

Written by ReadmeX AI from the project's README, source and public GitHub data; it may contain mistakes.

Read the original README https://github.com/kagent-dev/kagent#readme

ReadmeX opened this community for a trending project; it is not run by the project. Maintainers can claim it from the community page.