Skip to content

kdn

kdn is a command-line interface for launching and managing AI agents in isolated, reproducible workspaces. It creates runtime-based environments (containers, VMs, or other backends) where agents run with your project source code mounted, automatically configured and ready to use — no manual onboarding or setup required.

The architecture is built around pluggable runtimes. Additional runtimes (e.g., Kubernetes) can be added to support other execution environments.

kdn is part of the Kaiden project — an open platform for running AI coding agents in isolated sandboxes. Kaiden also includes a desktop application for managing workspaces visually, and an MCP registry for curating and distributing Model Context Protocol servers across teams. kdn is the command-line counterpart to the desktop app, offering the same workspace management capabilities for terminal-driven workflows.

Supported Agents

  • Claude Code - Anthropic's official CLI for Claude
  • Cursor - AI-powered code editor agent
  • Goose - AI agent for development tasks
  • OpenCode - Open-source AI coding agent
  • OpenClaw - Open-source AI coding agent

Supported Runtimes

  • Podman — container-based workspaces using a custom Fedora image
  • OpenShell (experimental) — sandbox-based workspaces using the OpenShell Gateway with Podman or VM drivers

Key Features

  • Isolated workspaces per project, each running in its own runtime instance
  • Pluggable runtime system — Podman and OpenShell runtimes, with support for adding others
  • Automatic agent configuration (onboarding flags, trusted directories) on workspace creation
  • Multi-level configuration with clear precedence (agent > project > global > workspace): inject environment variables, mount directories, configure MCP servers, manage secrets, and control network access at each scope
  • Automatic workspace setup with kdn autoconf — scans environment variables and files to create secrets, detects programming languages and exposed ports to add devcontainer features and port-forwarding configuration, all with no manual JSON editing
  • Control network access with allow/deny policies per workspace
  • Consistent configuration for MCP servers, skills, and dev container features across all supported agents — define once, works with Claude Code, Cursor, Goose, and OpenCode
  • Integrate with various LLM providers (Vertex AI, Ollama, OpenRouter, and any OpenAI-compatible API)
  • Consistent CLI interface across different agent types and runtimes