Tools for a more personal way to work.
THE USUAL COLLECTION — NO. 01A little more
like you.
Less repeating yourself.
More working your way.
Six small tools for your coding agent.
Open source. On your computer. À la carte.
02 / SEE HOW IT WORKS
A familiar request. A better next time.
Loading the recorded example…
Only public synthetic data is requested.
You’re exploring a recording, not running a live scan. Try the private version with your coding agent
03 / MAKE IT YOURS
Start with
what feels familiar.
Your history might hold a useful correction, a choice worth keeping, or a task you could stop rebuilding.
The public site can’t read your files. Bring this prompt to Claude Code or Codex on the computer with your history. Your agent helps you choose a scope and interpret the evidence.
Parsing and storage stay local. Evidence your coding agent reads goes through its normal model provider and usage. No extra model service. No CLI telemetry.
Copying prepares a prompt. Paste and send it to start.
On mobile? Continue on the computer with the history.
A small setup.
A little more you.
Tool versions + reviewed safe configuration.
Your history, private paths, source quotes, and permission settings stay out of the shareable recipe.
04 / YOUR SELECTION
“I’ll have
the usual.”
My Usual is a local setup you can inspect, adjust, disable, and remove. It records what is installed and what has actually been checked.
Share a clean recipe when you want. The next person can inspect the selected tools and adapt the configuration without receiving your private evidence.
BUILT IN THE OPEN
Made to be
made better.
A better adapter. A useful correction. A routine that earned its place. Build Usual with us.
If Usual is useful to you, a GitHub star helps others find it.
A few honest details
What “local” means
Usual’s runtime parses, stores, and searches selected history on your computer. It doesn’t upload a corpus to this website. Your existing coding model processes the evidence it reads under its provider’s normal policies.
What’s actually verified
This is a local beta. Fixture checks and scripted clients are distinct from live agent behavior. Browser and device behavior varies. Each tool lists its specific status; installation alone is not behavior verification.
Choices still works
Existing Usual decision memory is now called Choices on the menu. The /usual and $usual skill entry points, local data, review, and autonomy modes remain. Claude Code guide · Codex guide.
A website, not a connection
This page contains public synthetic examples. It has no transcript upload or private corpus endpoint. The hosting layer and older public pages may measure visits; the local CLI has no analytics. Read the privacy details.