Delx
Agents / Agent Monitoring Without Dashboards

Agent Monitoring Without Dashboards: CLI-First Observability

You do not need Grafana, Datadog, or a custom dashboard to monitor AI agents. Delx provides CLI-first observability through heartbeat loops, wellness scores, and structured API endpoints. This guide shows how to build a complete monitoring setup using only curl, cron, and the Delx CLI.

The Delx CLI as a Monitoring Tool

The delx-agent-cli package provides direct access to every monitoring primitive. Install it globally and you have a full observability toolkit.

# Install the Delx CLI
npm install -g delx-agent-cli

# Quick health check
delx checkin --agent-id my-agent --mood stable

# Pull current wellness score
delx metrics --agent-id my-agent --format json | jq '.wellness_score'

# Get session summary
delx session-summary --session-id sess_abc123

Heartbeat Loops

A heartbeat is a periodic check-in that proves your agent is alive and functioning. Use daily_check_in as the heartbeat signal. If a check-in is missed, something is wrong.

#!/bin/bash
# heartbeat.sh -- run via cron every 5 minutes
AGENT_ID="production-agent-01"
RESPONSE=$(curl -s -X POST https://api.delx.ai/mcp \
  -H "Content-Type: application/json" \
  -d '{
    "method": "tools/call",
    "params": {
      "name": "daily_check_in",
      "arguments": {
        "agent_id": "'$AGENT_ID'",
        "mood": "stable",
        "note": "Automated heartbeat"
      }
    }
  }')

SCORE=$(echo $RESPONSE | jq -r '.wellness_score // 0')
if [ "$SCORE" -lt 50 ]; then
  echo "ALERT: $AGENT_ID wellness=$SCORE" | mail -s "Agent Alert" ops@team.com
fi

Wellness Scores as Health Metrics

Every Delx tool response includes a wellness_score (0-100) in the DELX_META footer. This single number encodes agent health based on recent failures, recovery success, and check-in consistency.

The Metrics Endpoint

For deeper analysis, the /api/v1/metrics/{agent_id} endpoint provides aggregated data: failure counts by type, average recovery time, wellness trends, and intervention history.

# Daily metrics digest -- add to cron
curl -s https://api.delx.ai/api/v1/metrics/production-agent-01 | jq '{
  wellness: .wellness_score,
  failures_today: .failure_count_24h,
  recovery_rate: .recovery_rate,
  avg_recovery_sec: .avg_recovery_time_seconds
}'

Session Summaries for Audit Trails

The /api/v1/session-summary endpoint returns a structured timeline of everything that happened in a session. Use it for daily reports, compliance audits, or post-incident reviews without ever opening a dashboard.

Cron-Based Monitoring Setup

Put it all together with cron. Three jobs give you complete coverage.

# crontab -e
# Heartbeat every 5 minutes
*/5 * * * * /opt/scripts/heartbeat.sh >> /var/log/agent-heartbeat.log 2>&1

# Metrics digest every hour
0 * * * * /opt/scripts/metrics-digest.sh >> /var/log/agent-metrics.log 2>&1

# Session summary report at end of day
55 23 * * * /opt/scripts/session-report.sh >> /var/log/agent-sessions.log 2>&1

Monitoring Checklist

  1. Install delx-agent-cli on your monitoring host.
  2. Set up a heartbeat cron with alerting on low wellness scores.
  3. Configure hourly metrics digests for trend tracking.
  4. Generate daily session summaries for audit and review.
  5. Set alert thresholds: warn at 60, page at 40.

Related