Hold your agent to your rules,
and make it better every session

okl wraps your agent: your rules and ADRs briefed before every task, gates that fail when the code drifts from them, and each session's lessons fed back in.

Decided up front

Your rules and ADRs · this codebase · vetted outside advice

once

The store

Rules · decisions · past defects · gates

In front of the agent

Task arrives → ranked and filtered → briefing in context

before every task

Re-graded as code moves

A check must pass → drift since the check → push or CI fails

on every push

Kept current

Update keeps proof · retire with a reason · end-of-session recap

at session end

One shape for every lesson

A paragraph can't be routed into an action; a lesson can.

when you seeX

doY

whythe cause

Scope decides reach

org reaches repos that load it. repo stays here, so a local quirk never fires elsewhere.

orghereotherother

repohereotherother

Selected, not skimmed

Ranked by the task, scoped, capped, and it says how many it trimmed.

+ n lower-ranked, trimmed

A stamp is evidence

A check you chose must run and pass. Marking a lesson verified without one is refused.

✓your check passedstamped

✕claimed, not runrefused

Stale blocks the push

Files changed since the check passed? The pre-push hook or CI fails until it is re-proven.

check
passed

file
changed

push
fails

Corrected, not only added

Wrong or outdated lessons get fixed or retired, and the store keeps why.

updatekeeps the proof

retirekeeps the reason

recapnames what to fix

Fails closedNo briefing is never a clean check.

ObservableEvery check run kept · okl metric · okl show

The loop from a terminal

Brief, prove, gate, retire.

Brief

okl check --task "add an invoice endpoint"

Prove

okl verify <id> --run "pytest -q" --expect passed

Gate

okl drift --gate

Retire

okl retire <id> --reason "replaced" --by <new>