🧠 Inspiration

Developers often submit Merge Requests (MRs) with missing descriptions or unfinished TODO tags, slowing down the team. I wanted to move beyond "static scripts" and build an AI-driven assistant using the GitLab Duo Agent Platform that actually "understands" the context of a contribution.

⚙️ What it does

Smart MR Agent is an AI-powered gatekeeper. It doesn't just scan text; it uses the GitLab Duo ecosystem to evaluate if an MR is truly "Ready for Review." It checks for missing documentation, lingering TODO items, and provides conversational feedback to the developer directly within the GitLab workflow.

🛠️ How I built it

The project is built on the GitLab Duo Agent Platform. We developed a Custom AI Agent that triggers on MR events. We integrated a Python-based Tool that the Agent calls to perform deep-scanning of the MR metadata and diffs. By leveraging GitLab Duo’s prompt capabilities, the agent can decide whether a TODO is just a comment or a critical blocker that should prevent merging.

⚠️ Challenges we ran into

The biggest challenge was moving from a traditional "CI-script" mindset to an "Agentic" workflow. We had to learn how to register our Python logic as a Tool that the GitLab Duo Agent could autonomously decide to use.

🏆 Accomplishments that I am proud of

We successfully moved a manual "check-list" process into an autonomous AI workflow. The agent doesn't just fail a build; it explains why the MR isn't ready and suggests how to fix it, acting like a digital teammate.

🚀 What's next for Smart MR Agent

Anthropic Integration: Using Claude via GitLab Duo for even deeper semantic analysis of code changes. Auto-Fix: Allowing the Agent to automatically remove resolved TODOs or draft the missing descriptions based on the code changes.

Built With

Share this project:

Updates