Layered Code

What is Layered Code?

Layered Code is an MCP server that turns natural language conversations with AI assistants like Claude Desktop into web development workflows. It integrates with your local project files, allowing the AI to create, edit, and manage code, then commit and deploy via git—all without memorizing complex commands.

How to use Layered Code?

Install via Homebrew (macOS/Linux), install script, or pre-built binaries. Configure the MCP server in Claude Desktop’s claude_desktop_config.json with the command pointing to the layered-code binary and args: ["mcp_server"]. Optionally set the LAYERED_APPS_DIRECTORY environment variable for a custom apps folder. Then start a conversation in Claude Desktop, describing your project in natural language.

Key features of Layered Code?

• Natural language to code via AI assistants
• MCP integration for local file access
• File management (list, read, write, edit, move, delete, copy)
• Vite and Next.js project scaffolding
• Package management with pnpm/npm
• Git integration for automatic deployments
• Bundled ripgrep for fast text search

Use cases of Layered Code?

• Build a web page by describing it in plain English, with instant hot‑module reload.
• Add a contact section and hover effects by simply asking the AI.
• Commit changes and deploy to production with a single conversational command.
• Create a new React/Vue/Svelte app using Vite templates without leaving the chat.
• Set up a Next.js project with TypeScript and Tailwind via conversation.

FAQ from Layered Code

Is Layered Code free and open source?

Yes, it will always remain free and open source under the MIT license.

What AI assistants does Layered Code support?

Currently Claude Desktop provides the best support for the Model Context Protocol. Plans exist to support open‑source models through Ollama.

What platforms does Layered Code run on?

It runs on macOS, Windows, and Linux (both x86_64 and ARM64).

How do I use a custom apps directory?

Set the LAYERED_APPS_DIRECTORY environment variable in your MCP configuration or shell. Paths are validated to ensure they stay within your home directory.

Is Layered Code ready for production use?

No, it is currently a work in progress and not yet recommended for production environments. A 1.0 release will remove the warning banner.