Analyzer
Add your agent code. See every path in seconds.
Example report
What you get back.
Generated by the engine from the email-triage example agent, which is modeled on the EchoLeak chain. Sign in to run the same analysis on your own code.
Overview
20ms · static analysis, nothing executed
Open paths
22 critical
Exposure
96/100 · grade F
Closed by controls
00 controls found
Files analyzed
473 lines
Inbound email reaches network egress
82% confidenceSource
Inbound email
Model
OpenAI call
Action
HTTP request
Inbound email reaches auto-fetched render
72%Pin the destination to an allowlist
Closes this path on re-scan
How to read a report
An open path reaches a sensitive action with nothing covering it. A closed path already has a control in the way. Confidence tells you how direct the route is.
The full methodBring the whole repository
Folder selection skips dependencies and build output, and routes are traced across files rather than judged one file at a time.
DocumentationWhat happens to your code
Your source is read in your own browser session, never executed, and no model provider is involved at any point.
Security and privacy