Observability

Built-in metrics, dashboards, and log aggregation for full-stack visibility into your AI agent security posture.

Built-in Dashboards

ShieldAgent includes pre-configured dashboards accessible from the main navigation. All metrics update in real time with no additional setup required.

Proxy Overview

Request rate, latency percentiles (p50/p95/p99), outcome breakdown (allowed/blocked/shadow), and proxy health across all tenants.

Security Events

Injection detections, policy violations, anomaly scores, and risk threshold breaches plotted over time with agent-level breakdown.

Agent Activity

Per-agent tool usage patterns, request volume trends, risk score history, and behavioral baselines.

Audit Explorer

Full-text search across audit events with filters by agent, tool, outcome, and time range.

Compliance

Framework coverage scores, gap analysis results, and audit readiness indicators.

Key Metrics

ShieldAgent tracks and exposes the following metrics through the dashboard and API:

Total requestscounter

Total intercepted MCP requests by outcome (allow/block/shadow)

Proxy latencyhistogram

End-to-end proxy latency in milliseconds (p50/p95/p99)

Policy decisionscounter

Policy decisions by action (allow/deny/shadow)

Risk scoresgauge

Current per-agent risk score with trend direction

Injection detectionscounter

Prompt injection detections by detection method

Anomaly eventscounter

Anomaly detection events by severity level

Alerts & Notifications

Configure alert rules from the dashboard to receive notifications when security events require attention:

▸Risk score exceeds a configured threshold for any agent

▸Prompt injection detected with high confidence

▸Policy violation rate spikes above baseline

▸Agent behavior deviates significantly from learned baseline

▸Tool drift detected (new tool not in approved manifest)

Audit Log Search

The audit explorer supports structured queries to investigate security events:

# Filter by outcome

outcome: shadow | allow | deny

# Filter by agent

agent: coding-agent

# Filter by tool

tool: read_file | write_file

# Filter by risk level

risk: high | critical

Data Export

Export audit data and metrics for external analysis or compliance reporting:

▸CSV/JSON export of audit events with configurable date ranges

▸Scheduled export to your S3 bucket (Enterprise plan)

▸Webhook integration for real-time event streaming

▸API access for programmatic data retrieval