Skip to content

Your AI, by the numbers. Your plan, tailored.

pickagent works out how much you use AI, tells you which plan fits you and gives you your Replay: your AI profile to show off.

No terminal needed: a guided prompt interviews you in ChatGPT, Claude or Gemini and estimates your usage.

See an example verdict →

Scanner for Claude Code and Codex

Coming soon
npx pickagent scan

It isn't on npm yet, so don't copy it. When it ships, it reads your local logs and measures your usage instead of estimating it. What it will read.

Demo
pickagent

Replay · October

230 h

of agent working for you

Your archetype
The Night Owl
83% of your work, between 21:00 and 03:00.
Verdict
Claude Max
0 known blocks. With Pro you would have hit the limit 4 times a month.

Rank · Legend · Example · simulated data

How it works

Tell us. Compare. Show off.

  1. 01

    Your usage

    Today, a guided prompt interviews you in your chat app and estimates how much you use AI. The terminal scanner, which measures your local logs, is coming soon.

  2. 02

    Profile

    You get a summary of your usage, never your conversations. Every figure says whether it was measured or estimated.

  3. 03

    Verdict

    The algorithm replays your week against every plan in your browser: when you would hit a limit, what you would pay through the API, which combination works out best.

  4. 04

    Replay

    Your AI profile as cards: archetype, rank, figures and verdict. Share it in one tap and others can compare it with theirs.

The result travels in the URL's # fragment, which the browser never sends to the server. No account, and nothing is saved unless you choose to.

Demo

Move the levers

This is how the verdict feels. Change how hard your agent works and see how many times you would run out of quota on each plan.

Agent hours per day6 h
Agents in parallel2
Days per week5
Intensity
When you work

Blocks per week

Example · simulated data
  • Claude Pro

    1

    Claude Pro would block you 1 time a week. The first one is on Thursday at 12:00. Your week asks for 134% of the weekly cap. Anthropic doesn't publish its 5-hour window.

  • Claude Max

    0

    Claude Max: no known blocks. Your week uses 27% of the weekly cap. Anthropic doesn't publish its 5-hour window.

  • Claude Ultra

    0

    Claude Ultra: no known blocks. Your week uses 13% of the weekly cap.

Example verdict: the first plan with no known blocks is Claude Max. Its 5-hour window isn't published: we can't rule out blocks.

See my verdict with these values

Opens the real verdict with a profile estimated from these levers. We don't measure anything.

Day with a blockDay with no known block (estimated)Day without agents

Computed by the same engine as the verdict, with the Claude plans in the catalog (verified on Nov 5, 2026) and a profile estimated from these levers. Hourly consumption comes from a small measured sample (Nov 7, 2026). Pro's weekly cap is a community estimate and the 5-hour window isn't published, so here only the weekly cap blocks. The time of day doesn't change how much of the cap you spend, but it does change when the block arrives.

Radar · recent changes

Plans no longer compete on price: they compete on limits

Each date is the day the change was announced, with a link to the provider's source. Checked on Nov 7, 2026.

Browse the plans without measuring anything

Prices and limits for every plan, with source and date, open to everyone. And if you want to know which one fits you, the verdict replays your week against each of them.

We'll tell you when the scanner ships

The terminal scanner isn't on npm yet. Sign up and we'll email you when it ships. Just one email, nothing more. Meanwhile, the verdict already works with the guided prompt.