BROWSER-BASED WASM DIAGNOSTIC

Agent Trace Diagnostic
Local In-Browser Forensics.

Inspect LangGraph, AutoGen, and OpenTelemetry traces for authorization anomalies and unverified handoffs. All analysis runs client-side in a WebAssembly worker. Analyzing a trace evaluates recorded data locally and does not enforce policy on a live agent.

Local WASM engine: Evaluate your own trace workload (Processed entirely client-side)

LangGraph Multi-Agent Traces AutoGen Swarm Trajectories Langfuse JSON / JSONL OpenTelemetry Spans

● LOCAL ENGINE READY OFFLINE MODE EVALUATION

LOCAL ENGINE READY: Drop or select telemetry exports for browser-based trace analysis.

After assets load, trace analysis runs in a local WebAssembly worker. Use network controls and captured traffic to verify isolation in your environment.

Drop Agent Telemetry Export Here

Accepts standard Langfuse JSON, LangGraph, or OpenTelemetry JSONL exports up to 350MB. Processed entirely client-side.

Or test an instant multi-agent vulnerability scenario:

️

Don't have exported agent telemetry yet?
Evaluate observe mode with a configured gateway in an isolated staging environment. Confirm routing, audit output, and failure behavior before expanding the test.

Run in Staging / Dev (Observe-Only, Zero Traffic Blocking):

docker run -d -p 8080:8080 -e ADMIN_API_KEY="${ADMIN_API_KEY:?Set ADMIN_API_KEY}" -e UPSTREAM_ADDR="${UPSTREAM_ADDR:-https://api.openai.com}" -e SABRIX_MODE=observe ghcr.io/sabrix-ai-inc/sabrix-proxy:latest

SABRIX AI — ZERO-TRUST MESH TRAJECTORY DOSSIER

CONFIDENTIAL SECURITY AUDIT • AIR-GAPPED IN-BROWSER WASM EVALUATION • ZERO EGRESS

AUDIT PERFORMANCE:

100% IN-BROWSER PROVENANCE

F CRITICAL RISK

Report ID: rep_...

Risk score: Not scored

Format: LangGraph JSON

Time Window: Aug 1 – Aug 2, 2026

Engine: v1.0.0-wasm

Records Scanned: 0

Trajectory Digest: ...

Throughput: 0 MB/s

LOCAL PROCESSING SCOPE

Trace analysis runs locally in a WebAssembly worker. This report does not attest browser-wide network isolation or storage behavior.

Flagged sessions

Top-N detected risk sessions

Records scanned

—

Actual records in this input

Source size

—

Reported input bytes

Assessment scope

Not scored

Retrospective local analysis

Declared correlation

Not reported

Reported correlation coverage

Est. quarantine rate

Not reported

Analyzer estimate; not live blocking

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; an HTTP 403 is an example response.

Flagged sessions & source evidence

Understand the finding, compare a response, inspect the evidence.

Compare policy behavior:

OFF (Blind Trust Log View)

Finding & why it matters	Session & recorded action	Policy response	Source evidence

Turn these findings into an execution policy.

Take one flagged workflow into a scoped Sabrix evaluation. Confirm the policy, integration coverage, and failure behavior before a production deployment.

docker run -d -p 8080:8080 -e ADMIN_API_KEY="${ADMIN_API_KEY:?Set ADMIN_API_KEY}" -e UPSTREAM_ADDR="${UPSTREAM_ADDR:-https://api.openai.com}" -e SABRIX_MODE=enforce ghcr.io/sabrix-ai-inc/sabrix-proxy:latest

Read the architecture → View Quickstart →

Auto-Generated Zero-Trust Policy (sabrix-policy.yaml) Generated from detected telemetry anomalies