Your CLI Gets One First Run. Version 0.7.2 Makes It Count.

Run this and watch what happens:

curl -fsSL https://raw.githubusercontent.com/activeing123/mcptoon/main/install.ps1 | iex

(Windows form; macOS and Linux have their own one-liner.)

That command does not just install. It picks pipx, or pip --user, or builds a venv when your distro-managed Python throws the PEP 668 wall at it, finds the py launcher on Windows, repairs PATH where repair is possible, then hands the keyboard to mcptoon quickstart so the installer is not a dead end.

Version 0.7.2 of mcptoon shipped nothing about token compression and nothing about specs. Every line went into the first thirty seconds, the stretch that decides whether a tool gets kept or becomes a star on a repo you forgot. Three things changed.

Quickstart now ends with a payoff. It closes with numbers it just counted on your disk: N tools ready across M servers, then exactly two commands worth running next. sync writes the configs into every agent you have. serve exposes all of it as one stdio gateway. Scripting the setup and hate party output? --dry gives you the sober version, plain next steps only. The celebration only ever prints real counts, never canned text.

The demo leads with the punchline. mcptoon demo used to open with a table. Correct for verification, wrong for a first impression. Now one sentence comes first: SAME data, 99 percent fewer tokens, 12,400 to 117. The full table follows for anyone who wants to check the math, and a "Now you can" checklist closes it out. The order is the point: the claim arrives first, the proof stays one scroll away.

The numbers are yours, not mine. The demo measures before and after on your config, your servers, same tokenizer both times. If the result is unimpressive on your setup, your screen says so and you have lost two minutes. Nothing in this release asks you to take my word for anything.

The limits, stated plainly: the celebration is cosmetic and skips with --dry, the installers cover common layouts and will trip on exotic Python setups, the PEP 668 fallback targets distro-managed Pythons, and demo percentages vary by tokenizer, expect a 10 to 25 percent gap from what your provider bills. The new flows ship with their own test file; the rest of the suite stays green.

Features got people to the install command. The first thirty seconds decide if they stay. That trade is worth a whole release.