What Makes NightMend Different

You've tried Grafana + Prometheus. You know Datadog. They tell you something broke. None of them fix it.

NightMend is the first open-source AI platform that doesn't just monitor — it heals:

1. AI Analyzes — DeepSeek reads logs, metrics, topology to find the real cause
2. AI Decides — Picks the right Runbook from 13 built-in auto-remediation scripts
3. AI Fixes — Executes the fix with safety checks and approval workflows
4. AI Learns — Same problems get resolved faster next time

Global First: World's first open-source monitoring platform with MCP (Model Context Protocol) integration — your AI coding assistant can query live production data directly.

Quickstart

Try Online (no install): demo.lchuangnet.com — [email protected] / demo123

Self-Host in 3 Steps:

git clone https://github.com/LinChuang2008/nightmend.git && cd nightmend
cp .env.example .env                    # Optional: add DeepSeek API key for live AI
docker compose up -d                    # Open http://localhost:3001

First registered account becomes admin. On first startup, the backend auto-creates tables, alert rules, and dashboard components.

Feature Comparison

Sweet Spot: Small-to-medium teams who want AI-powered ops without enterprise licensing costs.

Honest disclaimer: We're early stage. For mission-critical systems at scale, use proven solutions. For teams ready to experiment with AI ops, we're your best bet.

How It Works

  Alert Fires        AI Diagnosis          Auto-Fix              Resolved
  ┌──────────┐     ┌──────────────┐     ┌────────────────┐    ┌────────────┐
  │ Disk 95% │────>│ "Log rotation│────>│ log_rotation   │───>│ Disk 60%   │
  │ on prod  │     │  needed on   │     │ runbook starts │    │ Fixed in   │
  │ server   │     │  /var/log"   │     │ safely         │    │ 2 minutes  │
  └──────────┘     └──────────────┘     └────────────────┘    └────────────┘

13 Built-in Runbooks: disk_cleanup | service_restart | memory_pressure | log_rotation | zombie_killer | connection_reset | cpu_high | docker_cleanup | network_diag | mysql_health | redis_health | nginx_fix | swap_pressure

AI Runbook Generator: Describe a scenario in natural language, and AI generates an executable Runbook with safety checks — via /api/v1/ai/generate-runbook.

Prometheus AlertManager Bridge

Already running Prometheus? Add 3 lines to alertmanager.yml and get AI diagnosis on every alert:

receivers:
  - name: 'nightmend'
    webhook_configs:
      - url: 'http://your-nightmend:8001/api/v1/webhooks/alertmanager'
        http_config:
          authorization:
            type: Bearer
            credentials: 'YOUR_TOKEN'
route:
  receiver: 'nightmend'

What happens: Prometheus fires alert → NightMend receives it → AI analyzes root cause → diagnosis appears in real-time on the Demo page via SSE.

Two modes: Diagnosis-only (safe, read-only analysis) or Auto-remediation (AI picks and executes the right Runbook).

Screenshots

MCP Integration — Global Open Source First

Your AI assistant (Claude Code, Cursor) queries live production data via MCP:

# Enable in backend/.env
NIGHTMEND_MCP_ENABLED=true
NIGHTMEND_MCP_PORT=8003
NIGHTMEND_MCP_API_KEY=your-secret-token

Note: Authentication via NIGHTMEND_MCP_API_KEY is required in production.

5 MCP Tools: get_servers_health | get_alerts | search_logs | analyze_incident | get_topology

Ask your AI: "Show all critical alerts on prod-server-01" / "Analyze last night's CPU spike" / "Search for OOM errors in the past 2 hours"

PromQL Query Support

Query metrics using familiar PromQL syntax via API:

# Instant query
GET /api/v1/promql/query?query=nightmend_host_cpu_percent

# Range query
GET /api/v1/promql/query_range?query=avg(nightmend_host_cpu_percent)&start=...&end=...&step=5m

# Supported: rate(), avg(), sum(), min(), max(), count(), avg_over_time(), label matchers

Compatible with Prometheus HTTP API format for Grafana integration.

Agent — Cross-Platform Monitoring

The NightMend Agent collects system metrics, discovers services, and monitors databases. It runs on Linux, Windows/Windows Server, and macOS.

Linux:

pip install nightmend-agent
nightmend-agent run -c /etc/nightmend/agent.yaml

Windows (PowerShell):

.\scripts\install-windows-agent.ps1 -ServerUrl "http://your-server:8001" -Token "your-token"
.\scripts\install-windows-service.ps1   # Register as Windows Service

Installation

Prerequisites

• Docker 20+ & Docker Compose v2+
• 4 CPU / 8 GB RAM (build) / 2 GB RAM (runtime)

Environment Variables

See docs/installation.md for full guide.

Tech Stack

Documentation

Getting Started | Installation | User Guide | API Reference | Architecture | Contributing | Changelog

Contributing

We need contributors who understand alert fatigue firsthand. See CONTRIBUTING.md.

cp .env.example .env
docker compose -f docker-compose.dev.yml up -d
pip install -r requirements-dev.txt
cd frontend && npm install

Community

• GitHub Discussions
• Report a Bug
• Email: [email protected]