Stop observing. Start Sensing. Sentinel is the first high-performance, WASM-extensible observability framework purpose-built for the scale, thermal complexity, and power-density of modern GPU clusters.
Explore the Docs | Build a Skill
Watch Sentinel detect thermal throttling and automatically mitigate it via the Fan Control Skill in real-time.
Build a Skill in Python, JavaScript, TypeScript, Rust, Go, C++, C#, Java, or Kotlin. Compile to WASM and see it live instantly.
A unified, enterprise-grade interface for managing global GPU clusters, thermal topology, and autonomous security.
Experience the full power of Sentinel + Prometheus + Grafana with a single command.
# Launch the full stack: Sentinel + Prometheus + Grafana
docker compose -f examples/full-stack/docker-compose.yml up -dWhat happens?
- Sentinel starts sensing your host (GPU, CPU, Power).
- Prometheus begins scraping Sentinel every 5s.
- Grafana is provisioned with an automatic datasource.
- Visit: http://localhost:3000 (Admin/Admin) to see your real-time AI infra health.
Sentinel is not just a tool; it is a movement toward Autonomic AI Infrastructure. Read our clear vision for the next era of computing:
👉 Sentinel Vision 2026: The Sentience Layer
ESNODE Sentinel is a high-performance, GPU-aware observability agent purpose-built for the scale and thermal complexity of modern AI infrastructure. It transforms raw hardware telemetry into Actionable Sentience.
- Polyglot Skills System: Modular architecture. Build "Skills" in Python, JavaScript, TypeScript, Rust, Go, C++, C#, Java, or Kotlin. Drop them in as WASM binaries without recompiling the core.
- AIOps Intelligence: Autonomous Root Cause Analysis (RCA) and Predictive Maintenance built directly into the agent.
- Power-Aware Orchestration: Real-time PUE calculation and performance-per-watt scoring for green AI.
- Modern TUI: A premium, dark-mode terminal interface for instant infrastructure command and control.
Traditional agents are rigid. Sentinel is a living organism. Extensions are Polyglot—write in your language of choice and compile to WebAssembly.
| Skill | Category | Capability | Viral Factor |
|---|---|---|---|
| GPU-Sentience | Core | NVML/ROCm Deep-packet telemetry. | High |
| Eco-Flow | Energy | Real-time PUE & Token/Joule Efficiency. | Hype |
| IoT-Bridge | Protocol | MQTT/DNP3 integration for Liquid Cooling. | Essential |
| Risk-Oracle | AIOps | Predictive GPU Failure Scoring (0-100). | Magic |
Contribute a Skill: Get listed in the Community Gallery and earn exclusive Sentinel Sentinel status.
Launch Sentinel with a single command:
# Clone the repository
git clone https://github.com/ESNODE/sentinel.git
cd sentinel
# Build and run
cargo run --release --bin esnode-sentinelesnode-sentinel cliSentinel exports a rich Prometheus-compatible metrics stream at :9100/metrics and provides a real-time JSON status API at /status.
{
"healthy": true,
"load_avg_1m": 0.15,
"gpus": [
{
"id": "GPU 0",
"util": 88.5,
"power": 320.4,
"risk_score": 12.0
}
]
}Built for the scale of tomorrow. 🦅
ESNODE | Source Available BUSL-1.1 | Copyright (c) 2024 Estimatedstocks AB
ESNODE Sentinel is designed for production reliability:
- Zero-Dependency Core: Single static binary for easy deployment.
- WASM Isolation: Third-party skills run in a secure sandbox.
- Enterprise-Grade TUI: No browser needed for deep diagnostics.
- Security & Cloud Console: Full Security & Console Guide for HTTPS and SSO.
- Global Fleet Orchestration: Learn about MCP & Autonomous Tuning.
- Secure Automation: Extend Sentinel via WASM Skills & Performance Layers.
- Strategic Adoption: Read our Enterprise & Mega-Cap Strategy for large-scale fleet integration.
ESNODE Sentinel is source-available under the ESNODE BUSL-1.1 license.
- Free for everyone with fewer than 100 GPUs under management.
- Enterprise License required for large-scale GPU fleets and commercial redistribution.
See LICENSE for details.
Built with ❤️ by the ESNODE Team