Your agent should know when to sprint. And when to stop.

Limitcheck gives coding agents a live sense of their remaining runway, so they can decide how deeply to work before the limit decides for them.

Read the 30-second docs

npx --yes limitcheckone-time setup

agent planningcodex
01
incoming taskaudit and refactor the authentication layer
5h windowresets in 2h 16m
62%
weeklyresets in 4d
18%
agent decisionsplit the refactor into two checkpointed passes
01

The check happens before the expensive part.

The installed skill teaches the agent when to look. Large refactor, long build, multi-file migration—it checks first, without you having to remember.

task receivedimplement workspace permissions
limitcheckreading available windows
03
nextscope work from the lowest runway
02

The shortest-looking limit is not always the one that matters.

A healthy five-hour window can hide an exhausted week. The agent reads every window and plans from the lowest one.

normalized runwaylowest wins
codex / 5h62% left
codex / weekly18% left
planning signal
split work · checkpoint often
03

One install. The agents already on your machine learn the habit.

Codex, Claude Code, Cursor, OpenCode, CrocBot, and custom harnesses all get the same small protocol. Provider auth stays where it already lives.

detected surfacesinstalled
  • codexskill
  • claude codeskill
  • cursorrule
  • opencodeskill

That’s the whole product. Install it once, and let the agent decide how much work its runway can support.

See exactly how it works