-
-
home page
-
after analysis
-
after analysis
Inspiration
Developers spend a lot of time dealing with CI/CD failures, reading long logs, identifying issues, and manually fixing them. This slows down development and breaks workflow momentum. We wanted to explore how AI could move beyond just suggesting solutions and actually take action inside real development pipelines. That idea led to building an autonomous agent that can handle failures on behalf of developers.
What it does
Agitic is an AI powered agent that integrates with GitLab workflows to automatically handle CI/CD failures. When a pipeline fails, the system analyzes the error logs, identifies the root cause, generates a fix, and creates a merge request with the proposed solution. It can also retry intelligently if the issue is not resolved in the first attempt, making the process more autonomous and reliable.
How we built it
We built Agitic using Python with a Flask backend to handle requests and trigger the agent workflow. The system uses an AI model to analyze logs and generate fixes. GitLab APIs are used to create branches, commit changes, and open merge requests. The overall system follows an event driven approach where a failure triggers the agent, which then executes a sequence of actions from analysis to resolution.
Challenges I ran into
One of the main challenges was handling unstructured and varied CI error logs and making sure the AI could interpret them correctly. Another difficulty was integrating smoothly with GitLab APIs for automated commits and merge requests. Ensuring the generated fixes were meaningful and not just generic responses also required careful prompt design and testing.
Accomplishments that we're proud of
We successfully built a working end to end system that goes from failure detection to automated resolution. The agent is not limited to giving suggestions but actually performs actions inside a real development workflow. Creating a system that retries and improves its response added an extra level of autonomy that we are especially proud of.
What we learned
We learned how to design event driven systems powered by AI and how to connect intelligent models with real world developer tools. Working with APIs and automating workflows gave us deeper insight into DevOps practices. We also learned the importance of clear system design when building multi step AI driven processes.
What's next for Agitic
We plan to expand Agitic into a multi agent system where different agents handle testing, security, and deployment tasks collaboratively. We also want to improve the accuracy of fixes, support more complex errors, and integrate deeper into development workflows to make it a reliable digital teammate for developers.
Log in or sign up for Devpost to join the conversation.