Free Exposure Scan
What have your AI agents already given away?
Every coding agent on a machine keeps a full local record of what it was told, what it read, and what it ran. Frostbridge Exposure Scan reads those records on your machine and shows what is already exposed. Read-only. Local-only. Free.
- Finishes in under two minutes
- No account needed to scan
- Nothing uploaded, nothing changed
- Signed and notarized for macOS
What it reads
- Agent transcripts from Claude Code, Codex, Cursor, Copilot, Windsurf, and Gemini
- Agent permission and MCP configuration
- Which credential files exist on the machine
- Shell history, and browser history for AI sites if you opt in
What it never does
- Upload anything, anywhere
- Change a single file or setting
- Ask for admin rights or your password
- Show a full secret. Findings are masked
Three steps, one machine
The scan runs where the records live. Nothing is installed beyond the app itself, and nothing leaves the machine.
-
01
Download and open
Pick the build for your machine. The app opens to a plain statement of what it will read and what it will never do.
-
02
Run the scan
Agree to the terms, choose whether to include browser history counts, and run. The scan reads local agent records only and finishes in under two minutes.
-
03
Read the report
An exposure score out of 100, the findings behind it, and the fix for each. Save it as a PDF from the report itself.
What the report shows
The report reads like a briefing, not a log dump. Each section answers one question about this machine.
- Exposure score
- One number from 0 to 100 with a risk band, and where the risk sits by pillar.
- Findings and fixes
- Each finding with its severity, the masked evidence behind it, and the change that closes it.
- Where your data went
- Which tools and servers your agents reached from this machine, and what kind of access each one had.
- What it cost, by month
- Estimated spend by agent, model, and month, and the share of it that was avoidable.
- Hardening playbook
- The concrete changes that close the findings on this machine, collected in one place.
Free to scan. Free to keep.
The scan, the exposure score, and the top findings are free and need no sign-up. The complete report, with every finding, where your data went, cost by month, and the hardening playbook, unlocks with a verified work email.
What unlocking sends
Exactly four fields you type: name, title, company, and work email. A six-digit code arrives by email and confirms the address. Nothing from the scan leaves the machine, before or after unlocking.
Questions before you run it
Does the scan send anything to Frostbridge?
No. The scan reads local files and writes its report on the machine. The only network call the app makes is the optional unlock flow, and that carries only the four fields you type.
Does it need admin rights?
No. It runs as your own user and never asks for a password.
Will it change anything on my machine?
No. The scan is read-only. The only files it writes are its own report and its own settings.
Which agents does it cover?
Claude Code, Codex, Cursor, Copilot, Windsurf, and Gemini, along with the MCP configuration and permission settings each one keeps locally.
Which systems does it run on?
Windows, and macOS on Apple silicon. Each download has a SHA-256 checksum beside it. The macOS builds are signed with the Frostbridge AI, Inc. Apple Developer ID and notarized by Apple.
Can I run it on more than one machine?
Yes. Run it on any machine you are authorized to examine. For a fleet, book a demo and we will walk through what an organization-wide view looks like.
One machine shows the exposure. An AI Officer governs what comes next.
The scan reads what already happened on one machine. Frostbridge governs AI use, exposure, model choice, tool authority, and cost before intent becomes action.
Or email hi@frostbridge.ai