This preview is password protected.
First, the honest part: a web page cannot scan your computer. The browser sandbox forbids it, and any site that claims otherwise is guessing. What this page can do is real: guide a two minute self sweep, hand you the same collector our platform uses, and analyze the evidence bundle right here in your browser. Your bundle never leaves this device. Nothing is uploaded, ever.
Answer what you find. Employee built agents leave artifacts and rhythm: a workspace of fingerprint files, something that wakes them on a schedule, and a process that talks to a model provider. One sign is a candidate. Signs in combination are a verdict.
The collector is a small Python script, standard library only, the same one the platform's fleet intake uses. It copies only the fingerprint files (HEARTBEAT.md and its family), snapshots your processes, schedulers, and current connections, and writes a bundle folder. Read it before you run it: it is short and commented.
Download the collector Download the network sensor
python3 collector.py --root ~ --out ~/gsis-bundle python3 sensor.py # optional: census of your local network
The network sensor lists devices on the local network and flags the ones nobody can name. It runs on your machine for the same reason: a website cannot see your network either.
Choose the bundle folder the collector wrote (the folder containing endpoints), or just the files of one endpoint. The analysis below is a faithful port of the platform's artifact census scoring, running entirely in this page.
What this page cannot do: check your company's registry to see whether a found agent is declared. That reconciliation is step one of the Census Sprint.
Do not delete it and do not confront it alone. An undeclared agent is a governance finding, not automatically a firing offense: most employee agents start as someone trying to do their job faster. The Census Sprint turns findings into a registry, zones, and policy, in two weeks.
Start the Census Sprint