OpsPilot: Discord-Native AI On-Call Team

A multi-agent ops copilot that triages incidents and ships safe PRs in minutes. Automated incident response with human oversight.

🎥 Demo Video

Watch Demo

Website

🚀 Quick Start

Prerequisites

• Node.js 18+
• Docker Desktop
• Discord & GitHub accounts

Setup

# Clone & install
git clone <your-repo> opspilot
cd opspilot
npm install

# Start infrastructure
npm run docker:up
npm run db:init

# Configure environment
cp apps/orchestrator/.env.example apps/orchestrator/.env
cp apps/discord-bot/.env.example apps/discord-bot/.env
# Edit .env files with your credentials

# Build & run
npm run build
npm run dev

# Register Discord commands
npm run register:commands

---

🧩 System Architecture

flowchart TD
    A[💬 Discord Bot] <--> B[⚙️ ADK-TS Orchestrator]
    B <--> C[🧠 MCP Servers]

    A --> D[🖥️ User Interface]
    D --> E[🤖 Agents: Watcher, Triage, Fixer]
    E --> F[🔗 Integrations: GitHub, Jira, PagerDuty, Prometheus, Runbooks]

Loading

Core Services

• Orchestrator: AI agents (Watcher, Triage, Fixer) with memory
• Discord Bot: User interface for incident management
• MCP Servers: Tool integrations (GitHub, Jira, PagerDuty, Prometheus, Runbook)
• PostgreSQL: Incident storage + pgvector for semantic search
• Redis: Job queue and caching

What It Does

1. Alert Detection: Watcher agent monitors Prometheus alerts
2. Auto-Triage: Analyzes incidents, queries metrics, identifies root causes
3. Smart Fixes: Generates PRs using runbook knowledge with approval gates
4. Seamless Integration: Links incidents to Jira tickets, PagerDuty, GitHub PRs
5. Continuous Learning: Updates runbooks from resolved incidents

Running & Testing

# Start all services
npm run dev

# Or run separately
npm run dev:orchestrator
npm run dev:bot

# Test the system
npm run test:demo

The demo simulates real incidents and shows the full workflow from alert → triage → PR creation.

Built with ❤️ by Himanshu Soni — aka h30s