Product · Scale
Runs at scale
Bulk work is something a run can simply do. The rows stay where they are staged and the script orchestrates by name, so a night's synchronisation is a few lines that survive a pause, a throttle and a restart.
Records matched across systems
Every record a script creates in another system is mapped to its source id, in an integration store the platform keeps per environment — no spreadsheet of ids, no lookup table in a script. The next run finds the mapping and updates instead of creating; a deleted mapping makes the next run create again, on purpose.
The bulk sync recipes in the docs →Thousands of records a night
Records are read into a dataset a page at a time, mapped, enriched one lookup per row, checked for duplicates and written out in bounded pieces. A bad row is set aside with its key and the reason; the job continues. A pause keeps the place, so nothing is read or written twice.
Recipes in the docs →Large files never break your script
A script holds a handle, never the bytes. Files are read and written by the platform, in file stores with per-file, per-store and per-environment ceilings; a CSV of any size is emitted in chunks that never split a record, and a far end's line number joins back to the record that wrote it.
Files and file stores in the docs →A fair queue, and ceilings you can read
Work is queued fairly across tenants; one tenant's burst does not delay another's night job. Every tenant runs under ceilings — queued work, stored bytes, outbound calls and mails — and a refusal names the limit that was reached. When the work jams, an operator reads the lanes, holds a run, stops one, or replays a logged event.
When the work jams, in the docs →Your own users
The people who call your endpoints — staff or customers — are users of your environment, never of the platform. You invite them by mail through your own mail account, give them the roles your environment defines, and your script sees who called. A folder of files served at your app's address signs them in at your environment.
App users in the docs →