BigCommerce stores run wide: long catalogues with variants and modifiers, several sales channels, and prices that change by customer group. The useful agent work here is bulk work, so the gate that matters is a record count rather than a single record.
01Use
Each one is scoped to a job a person already does by hand across thousands of records. Each writes through the BigCommerce APIs, and each stops at a decision you told it to stop at.
Assigns products to categories, maintains price lists per customer group, and manages which products are visible on which channel.
Fills variant and modifier data, custom fields and search keywords across the catalogue, in one reviewed pass instead of a spreadsheet import.
Reconciles orders arriving from several channels, adds notes to the ones that need a follow-up, and prepares refunds for a person to release.
02Wire
No copy of your catalogue sitting in someone else's database. The agent reads and writes live BigCommerce records through the APIs your store already exposes.
Rate limits and batch sizes are not an afterthought here. They go into the cost and latency budget we agree before the build, because at catalogue scale they set how long a run takes and what it costs.
03Gate
An agent that writes one record at a time as it thinks is fine on a small store and dangerous on a large one. Here the agent finishes thinking first, shows you the size and shape of what it wants to do, and waits.
It reads the catalogue, works out every record it would touch, and holds the result. Nothing has been written to your store at this point.
How many records the batch would change, which fields move, and a sample of before and after rows so you can judge the run without reading all of it.
Approval sits at the batch. Any write touching more than the record count you set waits here, and every price list change waits here regardless of size.
The approved batch writes within the rate limits and batch sizes agreed for your store, and the whole run is traced, so you can see afterwards what changed and why.
04Asked
Bulk writes are staged, not streamed. The agent prepares the whole change set first, reports how many records it would touch and a sample diff, then waits. A person approves the batch rather than each row. Eval cases written from your own products run before every change, and every run is traced, so a bad plan is caught before it reaches the catalogue.
Yes. Channels is part of the working surface, so a change is written against the channel you name instead of the whole store. That keeps a fix on one storefront from leaking across the others, and it lets order operations tell which channel an order arrived from.
Price Lists V3 is wired in, so the agent can maintain pricing per customer group. We treat every price list change as a gated write in every case: the agent prepares it, a person approves it, and only then does it go live. No price list edit ships without that step.
The category reassignment, the missing modifier data, the price list nobody wants to touch. We will tell you what the agent would read, what it would write, and what count should stop it.
