kaiden

A desktop application for developers building AI assisted tooling: chat with multiple model providers, run MCP servers and agent flows, and manage containers and Kubernetes, on top of Podman Desktop.

Kaiden is a desktop application for working with AI model providers, MCP servers, agent flow runtimes, and container and Kubernetes environments. Its stated audience is developers and platform engineers building AI assisted tooling who want one app for model access, agent execution, and infrastructure. Chat covers Anthropic Claude, Google Gemini, Mistral, any OpenAI compatible endpoint, OpenShift AI, and local models served through Ollama. The provider list leans broad on purpose, so the same window can talk to hosted services and local models without switching tools. MCP servers and agent flow runtimes round out the rest of the surface.

Kaiden reuses Podman Desktop's main, renderer, and preload architecture and its extension system. The container and Kubernetes features come straight from that project, which is how the app inherits that infrastructure surface without rebuilding it. Building on Podman Desktop means the extension hooks are shared with that project rather than invented fresh.

Under the hood it is an Electron application written in TypeScript with a Svelte renderer, organized as a pnpm monorepo. The main process package handles the plugin system and registries for providers, containers, Kubernetes, MCP, flows, and chat, while the renderer package is the Svelte UI. A deeper walkthrough of the plugin system and extension lifecycle sits in an agents documentation file, which is where the architecture gets its real explanation. The monorepo keeps the main process and the renderer separate, a common Electron pattern for keeping the two worlds apart.

• Source 1 · Project repository
• Source 2 · README