An MCP server for Packkit — let AI agents (Claude Desktop, Cursor, Windsurf, …) scaffold modern npm packages, CLIs, HTTP services, and front-end apps as a native tool.
packkit_schema — every option, preset, and alias as JSON (call first to learn the interface).
packkit_preview — the files a config would generate, without writing anything.
packkit_scaffold — generate a project to disk (optional git init + install).
Related MCP server: npm-run-mcp-server
Add to your MCP client config (e.g. Claude Desktop's claude_desktop_config.json):
{
"mcpServers": {
"packkit": {
"command": "npx",
"args": ["-y", "packkit-mcp"]
}
}
}Then ask your agent things like "scaffold a React component library called ui with Storybook" or "preview a Hono service named api".
Also listed on the official MCP registry as io.github.PackkitJS/packkit-mcp and on Glama.
Publishing a new version touches npm, the official MCP registry and Glama — see RELEASING.md.
MIT © DanMat
Show the project structure a config would produce — the full file tree and stack summary — without writing anything. Use this to check the layout before committing to it, and to show the user what they are about to get. Cheap and side-effect free, so prefer it over scaffolding to a throwaway directory to see what happens.
Inputs: name (required), preset, options.
Generate a project to disk. Writes files under / (or ./), optionally runs git init, installs dependencies, and creates the GitHub repository. Call packkit_schema first if you have not already — picking the preset by guess is how projects end up the wrong shape. If the target directory already has files in it (an existing clone, for instance), pass merge: true rather than failing.
Inputs: name (required), git, merge, github, +5 more.
START HERE. Returns every Packkit option, preset, and shortcut alias as JSON. Call this before scaffolding so you pick a preset that matches what the user actually wants — presets range from single libraries to CLIs, HTTP services, SPAs, and full-stack monorepos, and guessing without reading them is the main cause of scaffolding the wrong shape.
Inputs: no named parameters.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 3 tools
Each tool maps to a distinct stage: schema discovers valid options, preview simulates the output, and scaffold performs the actual write. There is no functional overlap or ambiguity between the three.
All tools share the packkit_ prefix and a single lowercase word, giving a clear and predictable namespace. No mixed casing or separator styles muddy the pattern.
Three tools is on the lower end but ideal for a narrow scaffolding domain: discover, dry-run, and generate. Every tool has a clear role and none feels redundant.
The set covers the full scaffolding workflow: discovering presets/options, previewing the tree, and creating the project on disk. There are no obvious dead ends or missing operations for the stated scope.
Independent directory of agentic AI tools — search, compare & recommend via MCP. Read-only.
A registry of AI agent tools — MCP servers, APIs, CLIs, SDKs — kept current by automated ingestion.
Gives AI agents direct access to your Sanity projects and guidance for working with them
Package intelligence MCP for AI agents — 22 tools, 19 ecosystems, AGPL SDK, free.
