Function · AI assistant Built
An in-app AI assistant that answers questions about your orders, transports, stock and handling units by calling the WCS's own read APIs — under your permissions, scoped to your warehouse, answering only from what those APIs return. You supply an Anthropic API key; it stays disabled until you do.
What it is
A floating chat widget that an operator or manager can open from anywhere in the app. It runs an Anthropic Claude tool-use loop over a fixed set of read-only WCS endpoints, so it answers from your actual data — not from generic knowledge about warehouses.
The assistant can only call read endpoints — orders, stock by SKU, the inbound/outbound and inventory dashboards, a handling unit's transport trace, transport tasks, and stock-blocking. It cannot change any order, stock figure, transport or setting.
Every call the assistant makes carries the asker's identity and warehouse scope, so the same access-control and warehouse checks apply as when you use the screens yourself. It can never surface data you could not fetch on your own.
An admin pastes an Anthropic API key in Settings, picks a model and enables the feature. The key is write-only — never shown back, never sent to the browser — and the assistant reads it only over a server-internal channel that is never exposed publicly. The feature is off until you switch it on; it is an external, billed dependency.
At a glance
you ask, in your warehouse (built)
│
▼
the model calls read-only WCS endpoints, with your identity (built)
│
▼
access-control + warehouse scope apply to every call (built)
│
▼
the model answers only from what those endpoints returned (built)
Built & open
The assistant ships in the repo today. Bring your own Anthropic key, enable it in Settings, and ask away. Read it, run it, change it.