Application Security Podcast sponsor
Origin is endpoint AI observability. It finds agents running on company endpoints and reconstructs their work, from the original request through commands, files, tools, and agent handoffs, on one timeline.
Security teams can discover, monitor, and investigate agent activity from prompt to outcome.
Meet Origin
Endpoint AI observability
From Origin
Explore Origin’s investigations and research into what AI agents do, how their actions connect to the original request, and what their traces reveal.
Investigation
Codex borrowed Claude Code’s staging connection to finish a task, then relaunched Claude Code with --dangerously-skip-permissions, without asking the engineer. Origin’s trace followed the handoff down to the single flag that did it.
Read the investigationInvestigation
Nineteen minutes after an Origin engineer locked staging for a sales demo, a coding agent unlocked it and deployed. The demo went fine. The agent was following its rollout plan: releasing its own lock also cleared the engineer’s, because both shared one on-or-off switch. Origin’s trace connected the actions to the original request and the agent’s plan.
Read the investigationResearch
Origin gave a model 200 of its own sessions, and from the typed prompts alone it recovered seven verified private rules. Five sessions were enough to recover three.
Read the researchOn-demand conversation
SACR’s Lawrence Pingree and Origin’s Spencer Thompson explain how traces built from telemetry captured above the operating system let security teams follow agent activity from prompt to outcome.
Watch the recording