Understand code before it runs.
Inspect intent, enforce a fixed policy, execute once under containment, and see exactly what the evidence proves.
Demo samples
Built-in samples use Vultr analysis and a fresh restricted sandbox run. Public demo results are shown in Reports and are not saved to shared history.
Each sample demonstrates one bounded behavior in a fresh sandbox run.
Execution policy
Fixed by the server, not by the model or upload.
AI planning can ask questions, but cannot provide Docker commands, images, mounts, environment variables, or paths.
Agent activity
—Run summary
A concise view of what the agent found, what ran, and what remains unknown.
No report loaded.
Behavior in this run was observed but this is not a safety certification.
Results reflect one bounded sandbox run and available host evidence only.
Run details
What the agent did
A completed stage describes work performed, not a verified security finding.
The program did not run in the sandbox.
AI assessment: What the agent discovered
Predictions are model estimates, not confirmed findings.
What the comparison established
No observed behavior is proof that the submitted program is safe. Independent operation-level telemetry is limited.
Sandbox configuration checks
No sandbox configuration report was produced for this run.
Run the benign containment probes to collect host-side results.
Evidence
Traceable records collected by the host or reported by the submitted program.
Configuration checks do not prove that individual file, network, or process operations were blocked.
Sandbox execution
Planned test cases
0 planned cases, 1 sandbox run. Planned questions are not separate executions.
Host metadata
Not available.
Program stdout
Not available.
Program stderr
Not available.
Run history
Authenticated records · newest first · source code is never stored.
Selected run
Select a saved analysis to inspect its report data.
No run selected.