How many tokens is my AGENTS.md?

Every line in AGENTS.md is paid again on every agent turn that loads it. A token count will not tell you whether a rule is useful — but it will tell you how expensive “always-on” instructions have become before you open Cursor, Claude Code, or Codex.

Fastest path: paste into a browser counter

1. Open the free AGENTS.md / CLAUDE.md token counter.
2. Paste the file (or drop text). Counting uses gpt-tokenizer with the o200k_base encoding entirely in your tab.
3. Read tokens, characters, lines, and rough input cost estimates for a few public list prices.

Nothing is uploaded. Treat non-OpenAI model counts as planning numbers — vendors differ slightly.

What “good” looks like

• Claude Code: documented target is under ~200 lines per CLAUDE.md; longer files cost more context and reduce adherence.
• Cursor: keep individual rules under ~500 lines; prefer scoped .mdc files over one giant always-on blob.
• Codex: combined project docs default-cap around 32 KiB and truncate past that.

If your root AGENTS.md alone is thousands of tokens, measure the rest of the stack too — Cursor rules, skills, and a parallel CLAUDE.md often double the bill.

When a counter is not enough

A single-file count misses duplicates and conflicting always/never lines across files. Use the full Clogmeter scanner when you want per-section bars, duplicate groups, conflict hints, and a trimmed export.