CI Whisperer

Testing Agent

Inspiration

We were inspired by the repetitive pain points developers face during code review and CI/CD troubleshooting. Manual reviews and debugging slow down teams and introduce risk. We wanted to automate these steps to help teams move faster and maintain high code quality.

What it does

CI Whisperer is an AI-powered agent for GitLab that: Reviews merge request diffs for bugs, risks, and missing tests Analyzes CI/CD job logs to explain failures and suggest fixes Generates concise release notes after merges Audits for security issues and unsafe configurations

How we built it

We built CI Whisperer in Python, using a modular architecture with separate components for review, CI analysis, docs, security, and fixes. The agent is orchestrated via YAML flows and responds to GitLab events. We focused on making the agent easy to extend and test locally.

Challenges we ran into

Designing a modular, event-driven agent architecture Integrating with GitLab events and CI/CD pipelines Generating useful, actionable, and minimal feedback Ensuring the agent’s suggestions are safe and non-destructive

Accomplishments that we're proud of

Accomplishments that we're proud of Automated the most tedious parts of the developer workflow Created a testable, extensible Python agent Provided clear, actionable feedback for code review and CI failures Enabled teams to move faster without sacrificing quality

What we learned

The value of automation in developer workflows How to design modular Python agents for real-world use The importance of concise, actionable feedback Best practices for integrating AI agents with CI/CD systems

What's next for CI Whisperer

Add deeper static analysis and test coverage checks Integrate with more CI/CD platforms Improve natural language explanations and suggestions Enable self-healing pipelines with automatic fixes