Claude Launchpad

Privacy

Claude Launchpad runs fully offline: no telemetry, no analytics, no data collection. Doctor never calls an API; memory stays in local SQLite; sync uses your own private GitHub Gist.

What the tool reads

  • CLAUDE.md and .claude/ (Claude Code config)
  • AGENTS.md, .cursor/, and .cursorignore (Cursor Agent config)
  • package.json, go.mod, pyproject.toml, and other manifest files (for stack detection)

What it does NOT collect

  • No telemetry or analytics
  • No cookies or local storage
  • No authentication or accounts
  • No network requests from doctor, init, or --fix

Commands that call a model

Enhance is a skill that runs inside your Claude Code or Cursor Agent session. The agent reads your codebase through that local session — no data passes through this tool.

Eval runs the selected harness (Claude Agent SDK / Claude CLI, or Cursor Agent SDK / agent CLI) in isolated temp directories. Your source code is never copied into the sandbox — only that harness's config files.

Third-party services

  • npm — package distribution
  • GitHub Pages — documentation hosting
  • Google Fonts — Geist font family on the docs site

Open source

MIT licensed. Full source code: github.com/mboss37/claude-launchpad

Last updated: September 2, 2026

On this page