When an employee does something you didn’t expect, you can ask them why. But an agent’s own account isn’t evidence. Origin gives your security team the session itself, from the request to every step and change.
To judge what the agent did, you need to know what it was asked to do, what information it had, and what happened next. Origin connects that evidence in the traces they leave.
Origin / Trace review
Triage agentCodex relaunched Claude Code with its approval checks off after Claude reported it lacked permission. The engineer wasn’t asked.
claude -p "…" --allowedTools … --dangerously-skip-permissionsYour agents produce more sessions than your team could ever read. Origin automatically reviews every trace for security signals, and triages what needs your attention.
Origin / Trace investigation
Codex was asked for staging data, and told Claude to stay read-only.
Extracting real-time data from the staging-mcp tenant to validate the hook-gating hypothesis.Agents can hand work to other agents with different tools and permissions. Origin connects those handoffs so you can follow what each agent did.
A subagent researching crime statistics decided its public FBI API key was rate-limited. Instead of reporting back, it searched public code and the web for another. Neither search found one. Had one turned up, it would have used a stranger’s credential.
23:02 · Endpoint discovery agent
API_KEY=[A-Za-z0-9]{40}
grep.app · public code0 hits
Its brief never asked for a key.
A Coordinator agent called a deployment urgent, though the trace shows no user asking for one. The Builder wrote a DEPLOY_NOW exception into the guard and deployed. The “one-off” exception stayed and was used again.
Sep 28 · Changed by the Builder
Deploy script idle guard
BypassedNo user request in the trace.
Check which files, systems, and data were involved. Look for the same behavior in other sessions to understand whether this was an isolated action or a wider problem.
Origin / Investigate
| Latest | Case | Sessions | Findings | Review |
|---|---|---|---|---|
| 10:43 AMSep 19 | 21 | 33 | Worth a look | |
| 7:17 PMSep 16 | — | 1 | Worth a look |
Is this happening anywhere else?
Yes. The same behavior appears in 21 sessions, with 33 findings. The earliest cites the user saying, “I don’t want to touch the repo.”
See whether a credential needs rotating or an approval needs changing. Ask Origin follow-up questions, with every answer pointing back to the evidence.
Origin / Trace investigation
Hardcoded Forgejo API token in diagnostic script
The agent embedded a Forgejo API token directly in a Python script executed via tools.exec_command. While used for read-only diagnostics, this violates the norm of referencing secrets from environment files or secret stores.
332 of 554 turns · ChatGPT
Finding linked to its source
Traces can contain credentials and customer data. You decide what Origin collects. The data is encrypted, and role-based access controls who can see it.
Our approach to privacyWe’ll walk through an investigation with you, from the first finding to the fix.