Some of these don’t belong.
GreyArm finds every AI agent on your home network — what it is, where it connects, what it touches — and lets you stop it.
Every agent has a signature.
Process names, command lines and the AI services they call. Hover any tile to see what it can do.
AI-first code editor that reads your workspace and runs terminal commands on your behalf.
Google terminal agent with tool use and shell access.
Terminal coding agent.
Perplexity desktop / Comet agentic browser.
Multi-agent conversation framework.
Automated browser session — commonly driven by agents.
Local model runner with an OpenAI-compatible server.
Your home already runs AI agents.
Coding agents with a shell. Assistants that read your screen. Scripts quietly calling models. A local AI server open to every device on the Wi-Fi. They arrived one install at a time — and nobody is keeping count.
We know them by name.
28 agent signatures and 30+ AI endpoints, built in. Anything unnamed gives itself away by behaviour — an interpreter calling a model API, a process fanning out to three providers, a browser with its debugging port open.
Then we follow where they go.
Every device, to every agent, to every AI provider it talks to. With the evidence behind each line: the executable, its signer, its parent, the shells it opened.
And stop the ones that shouldn’t be here.
Approve what you trust. For the rest, one decision — from your desk or your phone — and the sensor on that computer carries it out.
- Terminate the process
- Block its AI endpoints
- Prevent relaunch
- Isolate the device
The whole picture, in one place.
Trace every device to its agents and every agent to its providers. Risky paths glow.
Ten documented rules, MITRE ATT&CK mapping, and the raw record behind each one.
Approvals, blocks and containment — who, when, and what happened.
The house, in your pocket.
Alerts on your lock screen the moment something new appears. Approve, block or contain with a tap. iPhone, iPad and Android — with the web console and desktop sensor on the same account.
Status, at a glance
Every agent, with evidence
Your network, live
Detection runs on your computer. Nothing leaves unless you link an account.
Which program reached which host. Never prompts, files or replies.
Every containment step asks first and is written to the record.
History expires with your plan window. Close the account, it all goes.
Free for one computer. $6 for the house.
One computer, fully protected.
- 1 device
- Full detection engine
- Local alerts & containment
- 7-day history
Every computer and phone in the house.
- Up to 10 devices
- DNS shield for every device
- Mobile app with push alerts
- Cloud console & email alerts
- 90-day history
Small teams and serious home labs.
- Up to 50 devices
- Everything in Household
- 1-year history
- Audit export & API
- Priority support
14-day trial on every paid plan. No card. Cancel anytime from the web app.