What’s new in Sigil.

Routing rules can now weigh cost, latency and tool success rate together, and fall back across providers without code changes.

Agents no longer retry a timed-out tool call forever when the tool keeps returning partial results.

Search every span across every run by tool, model, cost or error, and jump straight to the step that failed.

Budgets now work per team and per agent, with warnings at 50%, 80% and 100% and an optional hard cap.

Agents deployed on Sigil Cloud now start 3.1× faster on average, with p95 cold starts down from 1.9 s to 480 ms.

Run the full Sigil control plane inside your own Kubernetes cluster with one Helm chart. Nothing leaves your VPC.