Sandbox
Dry run.
Paste an email body and see what the bot would do with it — which customer it resolves to, which jobs it extracts, and which orders it would create. Nothing is submitted to NICAM.
What it does
Runs the live pipeline
Same Claude extractor and grouping logic the real bot uses, against your live customer keyword map and validation rules from Settings.
What you see
Would-be orders
Customer match, extracted jobs, would-be PO groups, ref number, and a green “Would submit” or red “Would stop” badge with the exact stop reason.
What it doesn’t do
Never touches NICAM
No login, no order is created, nothing lands in the audit log. Costs one Claude API call per run — so be nice to the quota.
Result
Paste an email and click Run dry test to see what the bot would extract and how it would group jobs into orders.