Docs
Security and cost
The price tag on your agent's worst behavior. RepoOps already flags the insecure thing an agent did, and already knows what every session cost. This is the join: one row per session that tripped a security signal, with the dollar figure beside it, ranked by severity and then by cost. The session that was both risky and expensive sits at the top.
Where to find it
- Localhost:
/security-cost-fusion.html - API:
GET /api/security-cost-fusion - Sidebar: Guard, then Security and Cost
What it does for you
It answers the question that gets budget approved."An agent did something risky" competes with everything else on the list. "An agent did something risky and it cost this much" does not. The ranking puts the sessions worth a conversation at the top without you sorting anything.
One row per flagged session, with somewhere to go.Each row deep-links to the agent trace for that run and to the usage cost view, so the next question (what actually happened, and where did the money go) is one click rather than a search.
An unpriced session says unpriced.If a session used a model whose price is unknown, the row reads unpriced rather than showing a number derived from a guess. A fabricated cost on a security surface would be worse than a blank.
Built vs. planned
The join, the ranking, the deep links and the honest unpriced state all ship today. Both inputs are read live from disk, so nothing is cached into staleness. This adds no new capture and runs no LLM: it is the grouping of existing cost data by session that the telemetry surface does not expose on its own, joined to detections that already existed.
Last updated