Inspiration
Engineering teams spend a surprising amount of time manually reviewing, prioritizing, and organizing issues. As issue backlogs grow, it becomes harder to maintain consistency, identify duplicates, and determine what deserves immediate attention. We wanted to build an AI teammate that helps developers spend less time managing issues and more time solving them.
What it does
TriageBot is an AI-powered GitLab issue triage assistant. It analyzes GitLab issues, determines severity and priority, generates structured labels, identifies potential duplicates, recommends ownership, and can update GitLab issues with AI-generated triage assessments.
How we built it
TriageBot was built using Google ADK, Gemini 2.5 Flash, GitLab MCP, Docker, and Google Cloud Run. We connected Gemini to GitLab through MCP tools, allowing the agent to retrieve issues, analyze them, apply labels, create comments, and interact directly with GitLab projects.
Challenges we ran into
The biggest challenge was integrating GitLab MCP with a cloud-hosted agent. We had to solve authentication issues, configure environment variables correctly, create a custom Docker image with Node.js support, and ensure MCP tools worked reliably inside Cloud Run. Designing prompts that produced consistent triage decisions was another challenge.
Accomplishments that we're proud of
We're proud of building an agent that goes beyond answering questions and actually performs work inside GitLab. TriageBot can analyze issues, generate actionable insights, detect duplicates, recommend owners, and update issues automatically. We're also proud of successfully deploying the complete solution to Google Cloud Run.
What we learned
We learned how to build tool-using AI agents with Google ADK, connect external systems through MCP, deploy AI applications to Cloud Run, and design effective workflows that combine reasoning with real-world actions. We also gained a deeper understanding of the challenges engineering teams face when managing large issue backlogs.
What's next for TriageBot
Our next goal is to make TriageBot fully autonomous. Future versions will automatically triage newly created GitLab issues through event-driven workflows, assign issues to the most suitable engineers, provide team-wide backlog insights, generate daily triage summaries, and help engineering teams proactively manage project health at scale.
Log in or sign up for Devpost to join the conversation.