AutoMend AI Ops with Controlled Autonomy

Live Dashboard  ·  API Docs  ·  Pro Dashboard  ·  Changelog

---

Most infra tools tell you something is broken.
AutoMend investigates and fixes it.

Stop paging humans for disk cleanup, log rotation, and threshold alerts your monitoring already saw. AutoMend is an autonomous AI agent that monitors, diagnoses, and remediates system issues — then sends you a report.

Why this instead of scripts + alerts + DevOps tooling?

You already have…	AutoMend adds…
PagerDuty/OpsGenie alerts	An agent that reads the alert, runs diagnostics, and attempts a fix before waking you up
Cron scripts	Context-aware execution skips cleanup if disk is fine, escalates if a fix fails
Dashboards (Grafana, Datadog)	A companion that acts on what the dashboard shows, not just displays it
Runbooks	An agent that follows runbooks automatically and reports what happened

AutoMend doesn't replace your stack. It sits on top — watching, diagnosing, and fixing the things that don't need a human.

How it works: one real workflow

Disk usage hits 92% (warning threshold: 80%)
  ↓
AutoMend detects the alert
  ↓
Diagnostic Engine checks: log files, temp dirs, Docker overlay
  ↓
Identifies: unbounded nginx access logs (14GB)
  ↓
Auto-remediation: rotates logs, truncates old access logs, prunes Docker build cache
  ↓
Verifies: disk back to 54%
  ↓
Sends report to dashboard + email:
  "Disk usage spike resolved. Rotated nginx logs, freed 8.2GB."

All of this happens without a human in the loop — unless you configure it to require approval for destructive actions.

Safety boundaries

AutoMend is designed for controlled autonomy:

Action	Default behavior	Configurable
Read metrics & run diagnostics	✅ Always allowed	—
Log rotation	✅ Automatic	Can require approval
Temp file cleanup	✅ Automatic	Can require approval
Service restarts	❌ Requires approval	Can be enabled
Package updates	❌ Requires approval	Can be enabled
File deletion	✅ Age-gated (>24h)	Approval required for recent files
Config changes	❌ Requires approval	—

• Audit trail: every action is logged with timestamp, trigger, and result
• Rollback: remediation actions record state before changes where possible
• Sandboxing: destructive actions are dry-run first unless explicitly approved
• Kill switch: AUTO_REMEDIATION=false puts the agent in read-only mode

Dashboard

Demo Video

⬇ Download demo (48 MB, .mov)

Watch the full walkthrough: dashboard monitoring, real-time metrics, diagnostic analysis, and the Claude-powered chat interface.

Quick start

npm + pip

npm install
pip install flask gunicorn
cp .env.example .env
npm start                 # starts the agent
python3 dashboard/app.py  # starts the dashboard at http://127.0.0.1:8080

Docker

docker compose up -d
# Dashboard at http://127.0.0.1:8080

Architecture

index.js  (Node.js entry point)
├── src/agent/           MaintenanceAgent — orchestrates everything
├── src/monitor/         SystemMonitor — collects CPU, memory, disk, network
├── src/diagnostic/      DiagnosticEngine — analyzes health, suggests fixes
├── src/config/          ConfigManager — env-based configuration


maintenance_crew.py      CrewAI multi-agent system (optional)
dashboard/               Flask web dashboard (8 API endpoints)
.github/workflows/       Daily CI/CD at 3 AM UTC

Built with

Layer	Technology
Agent runtime	Node.js 20, native EventEmitter architecture
AI orchestration	CrewAI (optional multi-agent system)
Dashboard	Python 3.11, Flask, Gunicorn
Charts	Chart.js 4 with real-time data
Deployment	Vercel serverless Python
CI/CD	GitHub Actions (daily at 3 AM UTC)
Containerization	Docker, Docker Compose

Testing

pip install pytest
python3 -m pytest dashboard/

All endpoints return JSON. Data is deterministic within 5-minute windows (MD5-hash bucketed).

Configuration

Variable	Default	Description
MONITORING_INTERVAL	30000	Metrics collection interval (ms)
AUTO_REMEDIATION	true	Enable automatic fixes
NOTIFICATION_CHANNELS	email	Comma-separated channels
LOG_LEVEL	info	debug, info, warn, error

Alert thresholds (configurable via dashboard):

• CPU: warning 70%, critical 90%
• Memory: warning 75%, critical 95%
• Disk: warning 80%, critical 95%

Contributing

See CONTRIBUTING.md for how to get started. All contributions — bug reports, feature requests, docs, and PRs — are welcome.

License

ISC — see LICENSE.

Author

Built by kenGucci · @suggestionii