Patch

Patch is an AI-powered assistant for Jira that turns issues into pull requests with a single click.

Comment

Loom: https://www.loom.com/share/1acf3e235c7e4cf3857dcfb20d618f66?sid=a4efd7e7-7ce8-4f73-8531-bf7048c0352f

Youtube: https://www.youtube.com/watch?v=7Xll-Gjxmpw

Inspiration

As developers, we constantly switch between Jira and Bitbucket to track issues and implement fixes. We noticed how much time is wasted manually creating branches, commits, and pull requests — especially for repetitive or straightforward issues. We wanted to build something that closes that gap and brings AI deeper into the developer workflow.

What it does

When a user clicks the “AI Fix Issue” button inside Jira, Patch:

  1. Reads the issue title, description, and gathers repository context.
  2. Uses AI to generate the code patch or suggest changes.
  3. Automatically creates a pull request in the linked Bitbucket repository, referencing the Jira issue.
  4. Sends the developer a link to review, edit, and merge — all in one seamless flow.

This removes friction between project management and version control, turning ideas into implemented code instantly.

How we built it

Platform: Atlassian Forge Frontend: Custom Jira issueAction UI built with React (Forge UI kit) Backend: Forge resolver functions using Node.js (22.x runtime) APIs: Atlassian REST API for issue context + Bitbucket API for repo access AI Integration: OpenAI API for code and PR description generation Deployment: Secure Forge environment with external fetch permissions and OAuth integration

Challenges we ran into

  • Understanding the Atlassian Forge permission model — especially balancing jira:read and external fetch scopes
  • Debugging Forge deployment errors with manifest structures
  • Integrating OpenAI API calls securely through Forge without breaking authentication
  • Managing JSON responses between multiple APIs and ensuring consistent PR creation

Accomplishments that we're proud of

  • Building a working Forge app end-to-end in under 24 hours
  • Creating a foundation for AI-assisted software development directly inside the Atlassian ecosystem
  • Learning a lot about Forge’s capabilities and constraints (and how to work around them!)

What we learned

  • How Atlassian Forge apps integrate across multiple Atlassian products
  • The importance of permissions, authentication flows, and secure token management
  • How to use AI effectively for developer productivity rather than as a novelty
  • Collaboration and version control across multiple moving parts during a hackathon

What's next for Patch

  • Add support for Confluence — generating documentation updates automatically
  • Implement fine-tuned AI models for specific tech stacks or repos
  • Explore marketplace publication so teams can install Patch directly in their Jira Cloud

Built With

Share this project:

Updates