Chisle
The coding-agent plugin that bills 52% of a bare model across 20 live tasks. Terse persona + tool-output compressor + context diet.
翻訳待ち
リポジトリ情報はまだ収集していません。
No updates detected yet. GitHub releases will appear here.
macOS launchd scheduler that lightly primes Claude Code and Codex usage windows and logs quota snapshots.
翻訳待ち
Loading the full README.
Read the full README ↗Keep your Claude Code & Codex usage windows lit — on a schedule you choose.
Website · Download · Install · Configuration · Reference
▶ Watch the 40-second film · MP4
Many Claude and Codex plans limit how much you can use within a 5-hour window. The window starts with your first message and resets 5 hours later.
Say you start work at 9:00. Your limit resets at 14:00. If a tiny message had gone out at 7:00, it would reset at 12:00, and you would have a fresh window for the afternoon.
Stoker sends that tiny message for you, at the times you pick. This page calls it a check-in.
launchd fires the schedule. Nothing is sent to the author.READY. Codex reads one small file in
its own folder. Your real projects are never opened.You can use it from the terminal, from the menu bar app, or both. They share one engine and one config file.
Not every plan has a 5-hour window. Stoker reads the windows your account reports and recommends a schedule for them. As of November 2026:
| Tool | Plan | 5-hour window | Weekly limit |
|---|---|---|---|
| Claude | Free, Pro, Max, Team*, Enterprise (seat-based) | Yes | Yes, resets at a fixed time |
| Claude | Ultra | No (a temporary one until the account is verified) | Yes, resets at a fixed time |
| Codex | Plus, Business (standard seat) | Yes* | Yes |
| Codex | Pro, Business (premium seat)* | No | Yes, starts on first use* |
* Not confirmed by an official page. Plans change often, so treat the table as a guide.
A window starts when you first use it. A limit that resets at a fixed time does so whatever you do. What this means for you:
07:00, 12:00, 17:00, 22:00) keeps a window open through the day.To see what Stoker detected for your account, open the app's Settings or run
./install.sh recommend. Stoker only recommends. It never changes or turns off a schedule
by itself.
jq. The menu bar app includes it. For the command line, brew install jq.CLAUDE_STATUS_SOURCE=native.The Claude and ChatGPT desktop apps do not include these CLIs. Install them separately:
curl -fsSL https://claude.ai/install.sh | bash # then run `claude` once to sign in
curl -fsSL https://chatgpt.com/codex/install.sh | sh # then run `codex` once to sign in
Using only one tool? Set ACTIVATION_TOOL=claude or ACTIVATION_TOOL=codex.
stoker-gui-<version>.dmg from the
latest release.Stoker.app into Applications and open it from there.The schedule keeps running when the app is closed.
git clone https://github.com/hakupao/stoker.git
cd stoker
cp .env.example .env
./install.sh check # are the tools installed?
./install.sh dry-run # show what would be sent, send nothing
./install.sh install # load the schedule
Then set up sign-in for scheduled Claude runs.
By default a scheduled run reuses the sign-in from your own Claude Code sessions. That
sign-in is short-lived, so a scheduled run can fail with
401 Invalid authentication credentials. A long-lived token fixes this:
claude setup-token # opens a browser and prints sk-ant-oat01-...
Put the token in .env as CLAUDE_CODE_OAUTH_TOKEN="sk-ant-oat01-...". It uses your existing
subscription and adds no API charges. The app's Configure token button does all of this
for you.
Replace the app or pull the new code. Then run ./install.sh install once, or turn the
schedule off and on in the app. CHANGELOG.md says when a release needs more.
This is the beginning of the README. Read the rest at the original. Read the full README ↗