LOCAL TRACE AUDIT
Find risky actions in your agent traces—locally.
Inspect recorded tool calls, identify missing evidence, and export findings for a scoped enforcement evaluation—evaluated 100% client-side via WebAssembly.
Drop Agent Telemetry Export Here
JSON / JSONL up to 350 MB · Processed 100% locally in your browser sandbox.
v1 is optimized for native Langfuse JSON/JSONL and LangGraph exports. OTel OTLP format support is scheduled for v1.1.
SABRIX AI — LOCAL TRACE ANALYSIS REPORT
IN-BROWSER WEB WORKER EVALUATION • LOCAL CLIENT EXECUTION • NO REMOTE EGRESS
From trace evidence to execution control
See the risk behind the recorded action.
This is a retrospective audit with an illustrative policy simulation. No live request is blocked here. Actual enforcement requires a matching policy and a supported integration.
Flagged sessions & source evidence
Inspect the finding, compare simulated policy responses, and review evidence.| Finding & Impact | Session & Recorded Action | Policy Response | Source Evidence |
|---|
Turn these findings into an execution policy.
Take flagged agent workflows into a live runtime simulation or scoped Sabrix In-VPC evaluation. Confirm policies, approvals, and failure behavior before production deployment.
docker run -d -p 8080:8080 -e SABRIX_AUTHORITATIVE_TENANTS="acme" -e SABRIX_FAILURE_MODE=closed -e UPSTREAM_ADDR="${UPSTREAM_ADDR:-https://api.openai.com}" ghcr.io/sabrix-ai-inc/sabrix-proxy:latest
sabrix-policy.yaml)
Synthesized from detected telemetry anomalies
Illustrative starting template. Review the schema and adapt the rules to your tools and deployment before use.