Approval before changes
Anything that creates, changes, sends or deletes shows a preview and waits for a tap.
Every RageEdge assistant answers questions from your own data. If you want more, connect your tools and it can also check stock, create tickets and book slots, always within the rules you set. Connecting tools is optional.
Connected to this assistant
Example conversation
A name, what it does, the details it needs and whether it only reads or also writes.
A REST or GraphQL endpoint, a webhook, or a function in your own app through the SDK.
Which roles may use it, whether it needs approval, and what happens when the device is offline.
ai.registerTool( name = "book_slot", description = "Book a service slot for a vehicle", params = params("vin", "day", "time"), access = Access.WRITE, // READ or WRITE approval = Approval.ALWAYS, // ask before it changes data roles = listOf("service-advisor"), offline = Offline.QUEUE // send when back online) { args -> bookingApi.book(args) }
Code shown is illustrative.
Anything that creates, changes, sends or deletes shows a preview and waits for a tap.
Mark each tool as read only or write. Read tools never change anything in your systems.
A technician can check stock; only a service advisor can book a slot. Set it per role and per site.
Who asked, which tool ran, with which details, who approved it and the result, in the audit trail.
Changes made offline wait on the device, visible and cancellable, and are sent when it reconnects.
A new assistant answers only. Tools are added one at a time, and can be switched off at any moment.
No new middleware. If it has an API, a webhook or a function in your app, the assistant can use it.
No. Every assistant answers from your data without any tools. Add tools later, one at a time, when a team needs the assistant to act.
Only read tools you allow to run on their own. Anything that creates, changes, sends or deletes can be set to always ask first, and that’s the default.
Tools on the device, like the camera or files, work as normal. Tools that need your servers wait: changes are queued on the device, visible to the user, and sent when it reconnects. Answers from synced data keep working.
On the device. A small model picks the tool and fills in its details, so planning works without a network. Only the final call to your system needs a connection.
Every tool call lands in the console’s audit trail with the user, the details, the approval and the result. You can filter by tool, team or device.
Pick one team and one use case. We’ll have it running on your own devices in weeks.
Ask ChatGPT, Claude or Perplexity and see what they think. One click, instant insight.