// docs / tokens
Know what a prompt costs.
Two different meters matter: what your prompt costs every time an agent runs it, and what PLE's own AI features spend from your monthly budget. Both are visible.
The editor's ~token counter
Ln 1, Col 1 3 lines · 42 words · 291 chars · ~74 tok · markdownThe editor's status bar shows a live ~tokestimate as you type. It's the number to watch while trimming: a prompt that ships in every agent call pays its token cost on every call, so "same behavior, fewer tokens" is real money. Compare any two versions with diff to see what a rewrite actually saved.
Your monthly AI budget
PLE's own AI features, Improve and Loopvariant generation, spend from a monthly included-token budget (Pro 1M, Team 3M pooled). Each run reports what it used ("~N tokens used"), and Billing shows the running total for the period. The budget resets with your billing cycle. Running your variants through --exec is your own command on your own keys; PLE never spends your provider budget for you.
MicroBoost top-ups
Ran dry mid-month? MicroBoost($3) adds a one-time token pack on top of your included budget: bought from the Billing page, spent after your included tokens, and it doesn't expire with the cycle. No subscription change, no plan juggling.
What Free accounts see
The ~tok counter and version diffs are free for everyone; measuring costs nothing. The AI features that spend tokens (Improve, Loop) are Pro; the buttons say so instead of failing silently.