Order status for a customer question
When a customer emails about an order, the agent fetches the record at your request and drafts a reply with the current status.
Personal AI agent · REST APIs & webhooks
Connecting REST APIs & webhooks to your personal AI agent: if your own system or business software has a documented REST API or webhooks, we set up access on request. The agent reads first, webhooks notify it of events, and it only makes write calls after your approval.
Many companies run their own or specialized software, such as an order portal, a database of inspection reports or an industry application. With a documented REST API, the AI agent looks things up there instead of someone copying data by hand; webhooks tell it when something changes.
Open standard – connection on request
Connection on request. If your system has a documented REST API or webhooks, we check the documentation, version and permissions in the intro call, then set up the connection.
Tasks
When a customer emails about an order, the agent fetches the record at your request and drafts a reply with the current status.
When your system reports a new order or status change via webhook, the agent takes in the notification and prepares the next step for your approval.
On request, the agent pulls lists such as all open cases for a customer or all orders past their deadline and summarizes them.
If your software delivers documents through its API, such as inspection reports or delivery notes as PDF, the agent reads them and attaches them to a prepared reply.
If a status should be updated, a note added or a record created, the agent prepares the call, shows what would change and waits for your approval.
Workflow
01Read
The agent uses the approved read endpoints to fetch what a task needs, with its own access permissions.
02Prepare
It combines that data with mailbox, calendar and documents and prepares a reply, note or change with sources.
03Approve
For every write call, you first see which record would change and how, then decide: run, adjust or discard.
04Write / execute
After approval, the agent sends the call, and the audit log records what changed and who approved it.
Connection
REST is not a single product but a widely used way of building web APIs: data is retrieved and changed through fixed addresses, usually as JSON. We first read your system’s documentation, such as an OpenAPI description, clarify sign-in and permissions and begin with a few read endpoints.
Webhooks work the other way around: your system reports an event to the agent on its own. During setup, we check how it secures these notifications, for example with a signature. The agent responds with preparation, never with an unapproved change.
The agent starts with its own access, such as an API key, limited to reading the areas it needs. Write permissions follow only if you want write calls; even then it never sees more than the person it works for.
Anything that leaves your company or changes data is shown to you first – you send, edit or discard it. The audit log records what was proposed, when it was approved and which sources it relied on.
The agent runs on your own server in Germany or on your own hardware on your premises. For each request the AI model only receives the excerpt it needs; training on your data is excluded by contract.
Security & privacy on the product page →FAQ
In principle, any system with a documented REST API, from your own applications to business software. On request, we review documentation, sign-in and permissions and set up access, starting read-only.
The agent takes in the event, matches it to the right customer or case and prepares the next step, such as a reply or a note. Anything outbound or data-changing waits for your approval.
To the server your agent runs on: your own server in Germany or your own hardware on your premises. The AI model only receives the excerpt each request needs, and training on your data is excluded by contract.
Then we build a bridge: the agent operates the program like a person at the screen, in an isolated environment with its own user account, with every step logged. Writing steps come only once they run reliably, and only with your approval.
In an intro call we check your version and permissions, show the agent with mailbox and calendar, and plan the connection step by step.
Live presentation: we show the AI agent online on real workflows (in German) – free to attend. Register for the launch →
Trademarks belong to their owners. No partnership.