Kaiden

Kaiden is a desktop application for working with AI model providers, MCP (Model Context Protocol) servers, and container/Kubernetes environments from one workspace. It is built on the Podman Desktop extension architecture and adds AI-specific registries, ACP sessions, and RAG tooling on top of it.

What you can do with Kaiden

• Chat with multiple model providers — Anthropic Claude, Google Gemini, Mistral, any OpenAI-compatible endpoint, OpenShift AI, and local models served through Ollama or RamaLama.
• Connect MCP servers — install servers from registries and expose their tools to chats and agents.
• Manage containers — list, start, stop, build, and inspect containers, images, pods, volumes, and networks against Podman or Docker engines.
• Manage Kubernetes — switch contexts, view and edit deployments, services, ingresses, cron jobs, port-forward, and exec into pods.
• Build RAG pipelines — chunk documents with Docling and store embeddings in Milvus.
• Store credentials locally — provider API keys and tokens are kept in a local secret vault.

Who it is for

Kaiden is aimed at developers and platform engineers who are building AI-assisted tooling and want one application that covers model access, agent execution, and the container or Kubernetes environment the work runs in. It is also useful for evaluating multiple inference providers, including local ones, side by side.

How it relates to other tools

• Podman Desktop — Kaiden reuses Podman Desktop's main/renderer/preload architecture and extension system. The container and Kubernetes features come from that base; AI providers, MCP, and ACP sessions are added on top.
• LM Studio, Open WebUI, and other chat front-ends — those focus on local model inference and chat. Kaiden includes ACP sessions but also covers MCP and container/Kubernetes operations.
• Goose — Goose is an agent provider; Kaiden embeds it as one provider and adds a UI for managing model providers and MCP servers.

Architecture

Kaiden is an Electron application written in TypeScript with a Svelte renderer. The codebase is a pnpm monorepo:

• packages/main — Node.js main process, plugin system, and registries (ProviderRegistry, ContainerProviderRegistry, KubernetesClient, MCPManager).
• packages/renderer — Svelte UI.
• packages/preload, packages/preload-webview — IPC bridge between main and renderer.
• packages/extension-api — TypeScript API surface for extensions.
• extensions/ — built-in extensions: claude, gemini, mistral, openai-compatible, openshift-ai, ollama, ramalama, goose, mcp-registries, milvus, docling, container, kdn.

See AGENTS.md for a deeper walkthrough of the plugin system, IPC patterns, and extension lifecycle.

Status

Kaiden is at version 0.1.0-next. APIs, configuration formats, and the extension surface may change between releases.

Pre-built binaries for Windows, macOS, and Linux are available at openkaiden/prereleases. These are development snapshots built from the main branch and published automatically on each merge.

Prerequisites: Prepare your environment

You can develop on either: Windows, macOS or Linux.

Requirements:

• Node.js 22+
• pnpm v10.x (corepack enable pnpm)
• Gemini An API token for Gemini is required
• Goose Can be installed through the CLI panel

Step 1. Fork and clone Kaiden

Clone and fork the project.

Clone the repo using GitHub site:

git clone https://github.com/openkaiden/kaiden && cd kaiden

Step 2. Install dependencies

pnpm install

Step 3. Start in watch mode

Run the application in watch mode:

pnpm watch

Using Kaiden

Configure your Gemini API Key

Go to Settings > Resources, the Gemini provider should be listed. Click on Grab a key if you don't have an existing Gemini API key. If you already have one, enter it and click the Create button. If you go to the Chat window, you should see a list of Gemini models being listed.

Configure the GitHub MCP server

Click the MCP icon on the left toolbar. The Install tab should be selected and the GitHub MCP server should be listed. Click the Install Remote Server button on the left. Enter your GitHub Personal Access Token (PAT) if you have one or go to https://github.com/settings/personal-access-tokens/new to create a new one. Then click the Create button. If you go to the Chat window, you should see the server listed.

Contributing

See CONTRIBUTING.md for issue reporting, development workflow, and how to contribute extensions.

License

Apache-2.0. See LICENSE.