Bring agents onto your side.

Outside: see your public exposure

We agree authorised hostnames or public IP addresses and establish a baseline. The intended service checks DNS, HTTPS, certificates and visible configuration changes daily. Newly discovered hosts remain candidates until ownership and permission are confirmed. Your authorisation never extends automatically to suppliers, shared CDNs or neighbouring systems.

Inside: connect new reports with your software

An inventory or software bill of materials records your products and versions. We set up one agreed read-only import during the pilot, starting with a supported file format. New advisories can then be matched daily. A file import cannot automatically detect subsequent software changes: its collection date remains visible. A continuously connected customer agent requires later implementation and acceptance.

Published vulnerabilities and original research

Both packages follow newly published security information and relate it to your authorised systems. A possible version match, confirmed applicability and demonstrated exploitability are different conclusions. Findings retain their sources, timestamps and open questions. Finding a previously unknown vulnerability requires a separate research engagement; a zero-day discovery is not a promised outcome of monitoring. Active internal network scans and autonomous changes are outside these introductory packages.

A dashboard with the evidence behind each node

Explore agents, connected assets and findings in the interactive dashboard. Selecting a node reveals its task and result. The public demo is our own prototype with sample data; it does not monitor a real customer. Production data sources, model processing and alert delivery must pass end-to-end checks before operational service begins. The interface alone is not evidence that agents are connected.

How the connection will work

Hetzner hosts the dashboard, job control and evidence store. A model worker is intended to collect authorised analysis jobs through an outgoing connection and call a private, self-hosted model endpoint. A separate customer connector can later collect agreed inventory data. The model supports interpretation; an independent policy service limits targets and actions. Customer credentials and unrestricted scanning rights are not handed to the language model.

What makes the pilot complete

The 30-day observation period begins after the agreed technical acceptance. Human review and communication use the included monthly allowance during weekday business hours, 09:00–17:00 Europe/Berlin, excluding agreed local public holidays. Additional analysis or fixes require a separate order; alerts do not create insurance cover or guarantee that every vulnerability will be found.

1. 1Define: record authorised systems, permitted checks, contacts and a stop procedure.
2. 2Connect: establish the inventory or external checks and document the baseline.
3. 3Prove: process a real source update, demonstrate a clearly marked test finding and deliver a test alert.
4. 4Decide: review coverage and workload, then separately commission ongoing service.