For work worth repeating.Codex · Claude Code
Turn repeated AI work
into reusable code.
Clearings turns repeatable steps from your local agent's conversations into tested routines your agent can run again on fresh inputs.
Take it for a spinSmall routine.
Small routine.
Useful result.
Read the files. Find the exceptions. Leave the interpretation to your agent.
Change the receipt set.
Run it again.
Run it again.
research / receipt-auditBrowser-only sample
Check the latest research runs.
Show what needs my attention.
manifest.json8 runs
receipts/Example files
previous-summary.jsonExample baseline
Receipt audit is ready
Example dataRuns in your browser
Good work deserves
a second act.
One useful step. A few checked examples.
A tool your agent can reach for again.
From your conversation
Illustrative workflow“We keep checking
these receipts.
Make this reusable.”
Keep: collecting and checking files.
Leave: deciding what the findings mean.
Keep the thinking.
Lose the repetition.
Keep judgment with your agent. Save the collection, parsing, and checking that it should not need to rebuild.
- Good routine
- Audit research receipts and return the exceptions.
- Keep with the agent
- Decide what those exceptions mean for the research.
Still yours to change.
“Pause learning.”
You can pause, change the schedule, or undo an update through your coding client.
Explore the controlsExamples before activationChecked
Generated codeIsolated
File accessProject grants
Your routine libraryLocal