Follow sensitive data to where it went
One risky call is easy to spot. Data usually leaves in steps: a read, an archive, an upload. On Pro, advanced rules connect those calls and show them as one investigation in the hosted dashboard.
What it follows
Investigations cover four kinds of data.
Credentials and secrets
Credential files and recognized credential values, sent somewhere other than their own service. Reported by kind, never by value.
Database files and dumps
A database file or dump that is read, then sent to another host, as is or archived or encoded.
Mail stores
A local mail store that is read, then sent to another host, as is or archived or encoded.
Source code
Code pushed to a git host this machine has never reached before.
What one investigation shows
- Each call in the chain, in order, with its role: where the data came from, what was done to it, where it went
- The destination, and whether this machine had reached it before
- For a credential, its kind and whether it was sent as authentication or in a request, never its value
- Why the chain might be routine, so you can rule it out quickly
- Which interceptor saw each call
Clear about what it knows
The calls and their order are observed. That the data read is what was sent is inferred, because Omnodex sees the calls, not the bytes they moved. Each investigation says which is which.
Investigations explain what already happened. For activity through the MCP proxy they appear as the last call is recorded; for activity from agent hooks, when the session ends or within about 15 minutes. To stop a single risky call before it runs, use guardrails.
One investigation, step by step
Here a credential file is read, archived, then sent to a host this machine had never reached. Each step names the call behind it, and the investigation says why the chain might still be routine.
Where to find them
At the top of the Risk Events panel for the sessions in view, most severe first, and on the Investigations page, which lists every investigation across your sessions with filters for time range, severity, stream, machine, session, destination and rule. Like the rest of your data, they are decrypted in your browser.
Screenshots show generated demo data, not activity from a real user.