Kagent is an AI agent platform designed for Kubernetes environments, developed by Kagent Dev. It functions as an MCP server tool, exposing its running agents via an embedded Model Context Protocol server. This platform connects to various Large Language Model providers and integrates with external MCP servers, supporting advanced automation workflows. Kagent is developed using Go for its core components and also supports a Python ADK runtime for agents. It targets DevOps and platform engineers, enabling AI-driven operations within cloud-native infrastructures.

What does Kagent connect to?

Kagent connects to multiple Large Language Model (LLM) providers including OpenAI, Anthropic, Google Vertex AI, Azure OpenAI, Ollama, Gemini, and xAI. It also supports custom models. Beyond LLMs, Kagent integrates with external Model Context Protocol (MCP) servers. This allows Kagent to extend its capabilities by utilizing tools provided by other MCP servers, such as kagent-tool-server, which offers a rich set of Kubernetes-specific functionalities.

Core features and tools

Kagent's control plane includes an embedded MCP server exposing two core callable tools: list_agents, for discovering available agents, and invoke_agent, for running agents with specific inputs and supporting conversation sessions. Kagent also integrates with kagent-tool-server, which extends its capabilities with tools for Kubernetes (e.g., k8s_get_resources, k8s_describe_resource, k8s_get_pod_logs), Helm (e.g., helm_list_releases), Istio, Prometheus, Grafana, Cilium, and Argo.

Who is it for?

Kagent is primarily for DevOps and platform engineers. It enables them to automate infrastructure tasks, react to cluster events, and manage deployments within Kubernetes. The platform also supports incident response workflows, helping with alert triage and drafting runbooks. Developers can use it for platform self-service, requesting resources in natural language. This is not the right fit if you require Server-Sent Events (SSE) as a transport protocol, as it is not currently supported.

How to get started

For LLM providers, Kagent authenticates using API key authentication, typically stored in Kubernetes Secrets. For MCP servers, standard HTTP authentication methods, including Bearer tokens and API keys, are supported, with OAuth recommended for obtaining tokens.

Use cases

• Kubernetes Cluster Automation & DevOps: Automating infrastructure, reacting to cluster events, performing remediation, and managing deployments.
• Incident Response: Triaging alerts, correlating traces, drafting runbooks, and opening rollback pull requests.
• Platform Self-Service: Developers requesting resources like namespaces or databases in natural language, and agents filing GitOps pull requests.

Kagent serves as an AI agent platform for Kubernetes, developed by Kagent Dev, enabling automation and self-service through its MCP server. It supports various LLM providers and integrates with specialized tools for cloud-native operations. The platform offers capabilities for DevOps, incident response, and developer self-service. Try automating a Kubernetes infrastructure task using the invoke_agent tool.