Docs menu
Get started
Quickstart Install: Claude Code Install: Codex Install: other clients API keysConcepts
How a run works The app profile The severity model What a run does not sayTools
check_access claim_trial_key set_profile run_lint get_law resolve_domain_jurisdiction submit_feedback upload_lint_runGuides
A whole run, end to end Console TestPack in CIReference
Endpoint and transport lexlint.yml schema Errors ChangelogHelp
Support and feedbackGet started
Run LexLint once before you install anything
No install, no account and no key: paste one line into the agent session you already have open.
The first run needs no install, no account and no key. It happens inside a session you already have open: paste this line into it.
Your agent fetches the procedure, works out what your app does by reading the repository, takes a 30-day trial key with no sign-in, and reports what the lint found. It closes by uploading the run: you see the payload first, and the reply carries a link that opens the full run with no sign-in. What goes is the lint's own output, the two lists you declared and the repository's name; none of your code, schema or data is sent, and nothing in your working tree changes.
It runs in a session you drive, which is the point: your client asks before each command, and you see every endpoint it touches. There is deliberately no one-line unattended version. The procedure arrives over the network, and a command that pre-approves file access and outbound requests before anyone has read it would give that fetched text more trust than a first run has earned.