Skip to main content

🤖 Dual-MCP Architecture

LaraKube CLI is the first orchestrator to implement a Dual-MCP (Model Context Protocol) system. This provides AI agents with the perfect balance of "Local Reach" and "Global Memory."

🛠 1. The Local Mechanic (CLI MCP)​

The CLI binary acts as the Local Mechanic. It has direct access to your host machine's files and terminal.

  • Primary Context: Current working directory, source code, and local .env files.
  • Transport: stdio (Incredibly fast for local tools).
  • Best For: Cursor, Windsurf, and local terminal agents.
  • Capabilities: Modifying code, patching .env, and running orchestration verbs like up or add.

🧠 2. The Master Architect (Console MCP)​

The Console application acts as the Master Architect. It has access to the global project registry and history.

  • Primary Context: Project fleet status, historical audit logs, and remote cluster state.
  • Transport: SSE (Accessible via web URL).
  • Best For: Web-based agents like Gemini, Claude Web, or remote management tools.
  • Capabilities: Analyzing fleet health, debugging historical failures, and managing cross-project dependencies.

🤝 How they work together​

When you ask an AI agent to "Add Redis to this project," the Local Mechanic handles the file changes, while the Master Architect provides the historical context of why the previous attempt might have failed, ensuring a higher success rate.

Configuration​

To register these servers with your AI tools:

  • Local Agent: larakube mcp
  • Web/Remote Agent: https://console.kube/mcp