What it does
@goke/mcp bridges goke CLIs and MCP clients. It wraps your existing goke-based CLI and exposes each command as an MCP tool, making it discoverable and callable by Claude and other MCP clients. When you invoke createMcpAction({ cli }), every .command() definition becomes a full MCP tool, complete with argument validation inherited from your Zod schemas and help text automatically derived from your CLI definitions. The server handles argument parsing, type coercion, and tool discovery with no additional configuration.
Who it's for
TypeScript developers building internal or custom CLIs who want Claude to invoke them directly. DevOps and platform engineers with deployment, admin, or infrastructure-management tooling. Anyone maintaining a goke-based CLI who wants agent integration without writing a custom MCP server from scratch.
Common use cases
- Expose a TypeScript CLI to Claude Code so agents can run commands directly.
- Wire up internal deployment or database admin CLIs for agent access.
- Make existing goke tooling available to AI agents without custom server code.
- Add agent-driven automation to a command-line tool without modifying business logic.
Setup pitfalls
- The CLI can read and write the filesystem—sandbox accordingly in restricted environments.
- Requires a working goke CLI already built and installed; won't generate one from scratch.
- Zod schema options must clearly describe parameters so agents can discover and use them properly.
- No CI pipeline—verify the installed version works in your specific environment before relying on it.