SprintMind

SprintMind catches Friday sprint failures on Tuesday, a GitLab Duo Agent that watches your team every morning and tells you exactly what to fix before it becomes a missed deadline.

Comment

Inspiration

Every team finds out on Friday what went wrong on Tuesday. The signals were always there — a stale MR, a silent developer, an unreviewed PR — but nobody was watching. SprintMind was built to be that watcher.

What it does

SprintMind is an AI agent for GitLab that runs every morning via GitLab CI, analyzes your team’s merge requests, commits, and issues using Claude AI, and posts an actionable sprint health report as a GitLab issue. It highlights contributor status, top actions, burndown prediction, and a sprint rescue plan. It also comments on stale MRs and @mentions developers who need attention.

How we built it

A single Node.js file running inside GitLab CI on a daily schedule. It fetches data from the GitLab REST API, sends structured prompts to Claude Sonnet, parses the JSON response, and writes results back to GitLab as issues and MR comments.

Challenges we ran into

Getting Claude to return consistent JSON across different repo states. Building trend tracking without external storage by reading previous SprintMind issues. Preventing duplicate reports while keeping the agent re-runnable.

Accomplishments that we're proud of

A fully working agent in a single file. Proactive daily monitoring with no manual trigger. Real @mention notifications and multi-project support.

What we learned

GitLab CI is a powerful runtime for agents, and Claude excels at converting raw engineering data into clear, prioritized actions.

What's next for Untitled

Slack notifications, configurable rules, reviewer suggestions, and sprint velocity tracking.

Built With

  • anthropic
  • axios
  • claude
  • gitlab
  • gitlabapi
  • gitlabci
  • node.js
Share this project:

Updates