Ticketwatcher

• 

  Mermaid Graph Workflow

• 

  Quick Start Guide

Inspiration

We were inspired by the growing push to integrate agentic systems into real-world problem solving. As developers, we wanted a tool that could help fellow engineers manage their GitHub tickets—automating the repetitive parts so they could focus on the creative ones.

What it does

TicketWatcher is a library that automates the process of handling tickets relating to a GitHub codebase. It reads stack traces and hints directly from tickets, fetches the right code snippets, and orchestrates our agentic workflow to propose safe patches. The goal is to give developers instant insights into what broke, where it broke, and—when possible—open a draft PR with a proposed fix.

How we built it

Our workflow:
GitHub Issue → Webhook → TicketWatcher → AI Analysis → File Fetching → Fix Generation → Draft PR

Challenges we ran into

• Agentic orchestration: Balancing autonomy with developer oversight
  
• Design tradeoffs: Choosing between simplicity vs. flexibility in workflow
  
• Path resolution: Normalizing file paths across repositories and platforms
  
• Context limits: Keeping AI analysis both efficient and accurate
  

Accomplishments that we're proud of

• GitHub MCP for repository interactions and pull requests
  
• AI Orchestration to understand issues and generate fixes
  
• Path + Context Management for handling file lookups and large codebases
  
• Cloudflare to secure and streamline infrastructure
  

What we learned

• How to design and orchestrate agentic systems effectively
  
• The tradeoffs between context size, accuracy, and performance
  
• The importance of user experience when interacting with AI-driven tools
  

What's next for Ticketwatcher

• Support for custom agents
• Ticketing trend analysis
• Support for any git repo
• Optimization of workflow for better insights

Built With

• agentic
• cloudflare
• github
• javascript
• mcp
• openai
• python
• wrangler