AI Integration

Marmotte ships a built-in MCP server (marmotte-mcp) that lets any MCP-compatible AI host scaffold and configure projects without writing a single command. It also exposes a programmatic API for Node.js scripts and AI-generated code.

What is MCP? The Model Context Protocol is an open standard that lets AI assistants call external tools. Once a server is registered, the AI can invoke create or setup the same way it calls any other tool — no glue code needed.

---

Step 1 — Install marmotte globally

The marmotte-mcp binary is bundled inside the marmotte npm package. Installing it globally makes the server available to every AI host on your machine:

npm install -g marmotte

Verify it works:

marmotte-mcp --version   # should print the version and exit

No global install? You can use npx marmotte-mcp everywhere below instead. npx downloads and caches the package on first run.

---

Step 2 — Register the server in your AI host

Pick the host you use.

Claude Code (CLI)

Run this once — it writes to your user-level config so the server is available in every project:

claude mcp add --scope user marmotte -- marmotte-mcp

Verify it was added:

claude mcp list

See the Claude Code MCP docs for the full reference on scopes, environment variables, and project-level .mcp.json files.

Claude Desktop

Open the config file in a text editor:

Platform	Path
macOS	~/Library/Application Support/Claude/claude_desktop_config.json
Linux	~/.config/Claude/claude_desktop_config.json
Windows	%APPDATA%\Claude\claude_desktop_config.json

Add the marmotte entry inside mcpServers (create the file if it doesn't exist):

{
  "mcpServers": {
    "marmotte": {
      "command": "marmotte-mcp"
    }
  }
}

Then restart Claude Desktop. The server appears in the tool list on the next launch.

See the Claude Desktop MCP guide for screenshots and troubleshooting.

Cursor

Open or create .cursor/mcp.json in your home directory for a global setup:

# macOS / Linux
~/.cursor/mcp.json

# Windows
%USERPROFILE%\.cursor\mcp.json

{
  "mcpServers": {
    "marmotte": {
      "command": "marmotte-mcp"
    }
  }
}

For a project-scoped setup, place the file at .cursor/mcp.json in the project root instead.

See the Cursor MCP docs for the full reference.

---

Step 3 — Use the tools

Once the server is registered, ask your AI assistant to scaffold a project naturally:

"Create a new TypeScript library in ./packages/logger using the node-library template, no examples, with lint and format."

The assistant will call the create tool with the right inputs. No copy-pasting commands.

Available tools

create — scaffold a new project

Input	Type	Description
dir	string	Target directory, e.g. "./my-lib"
template	string	Template ID or source (see below)
name	string?	Package name. Defaults to directory basename.
examples	boolean?	Include example files (default: true). Built-in templates only.
features	string[]?	Feature IDs to apply after scaffolding.

Template values:

Value	Meaning
node-library	TypeScript library with Lib(), DTS, and docs
ui-library	Vue 3 + Vuetify component library
ui-app	Vue 3 + Vuetify SPA
user/repo	GitHub shorthand — downloaded via giget
github:user/repo	Explicit giget source
./path/to/template	Local directory

setup — add tooling to an existing project

Input	Type	Description
dir	string?	Project root. Defaults to current working directory.
features	string[]	Feature IDs to apply (at least one required).

Feature IDs: lint, format, pre-commit, changesets, version-lifecycle

Resources

The server also exposes two resources that the AI can read to discover valid values before calling a tool:

URI	Contents
marmotte://templates	JSON list of { id, label } for each built-in template
marmotte://features	JSON list of { id, label } for each available feature

---

Programmatic API

For scripts and AI-generated code that need to call the CLI logic directly, import from marmotte/cli/api:

import { runCreate, runSetup } from "marmotte/cli/api";

// Fully non-interactive when template is supplied
await runCreate({
  dir: "./my-lib",
  template: "node-library",
  name: "my-lib",
  examples: false,
  features: ["lint", "format"],
});

// Add features to an existing project
await runSetup({
  dir: "./packages/my-lib",
  features: ["lint", "format", "pre-commit"],
});

Omit any field to fall back to interactive prompts for that field. Supplying template in runCreate disables all prompts.

Discover templates and features at runtime

import { builtinTemplates, features } from "marmotte/cli/api";

builtinTemplates.map((t) => t.id);
// ["node-library", "ui-library", "ui-app"]

features.map((f) => f.id);
// ["lint", "format", "pre-commit", "changesets"]

Types

import type { CreateOptions, SetupOptions, Feature, PackageJson } from "marmotte/cli/api";

Type	Description
CreateOptions	Options for runCreate
SetupOptions	Options for runSetup
Feature	{ id, label, apply(dir, pkg) }
PackageJson	Minimal package.json shape used by features