GitHub Copilot for Jira gets custom agents, customized branching rules, and code review notifications

Custom agents and customization per Jira space

The announcement in the GitHub Changelog from April 22 brings features that move the Copilot cloud agent in Jira from the zone of a generic assistant toward a tool that can actually be calibrated to each team’s needs. The most important new feature is the ability to define custom agents — specialized Copilot configurations with their own system instructions and behavior rules. A team can have one agent for processing bug tickets that follows a strict reproduction and regression test protocol, another for preparing documentation, and a third for routine dependency updates.

Along with custom agents, GitHub has added support for Atlassian custom fields. This means that Copilot can read and interpret data from fields that the company itself has defined — for example a “Regulatory Impact” or “Client Type” field — and make decisions based on them about how to process the ticket. Before this update, Copilot relied only on standard Jira fields, which limited its usefulness in large organizations with developed metadata schemas.

Branching rules and conventions in large teams

The second significant new feature is customized branching rules at the Jira space level. Many companies have strict Git branch naming conventions — for example feature/JIRA-123-short-description, bugfix/JIRA-456-crash, or special prefixes for hotfixes. Previously, Copilot used built-in assumptions, which in practice meant that developers manually changed branch names after the agent opened them. Now a space administrator can define a branching rule that Copilot automatically follows when opening a new pull request linked to a ticket.

Along with branching rules, custom instructions at the space level are also available. These are text guidelines that Copilot respects in every interaction within that Jira space — for example “always include a regression test,” “follow the internal commit message style,” or “check CHANGELOG before merging.” The combination of custom agents, custom fields, branching rules, and space instructions makes Copilot a configurable tool that can genuinely fit into mature development processes.

Code review notifications — closing the loop with project management

The last announced feature may be the most practical for non-technical team members: code review request notifications arrive directly in Jira. When Copilot or a developer opens a pull request and requests a review, the notification appears on the linked Jira ticket. A project manager or product owner tracking the ticket in Jira can now see that a PR has been opened, who was asked to review it, and what status it is in — without needing to switch to GitHub.

This change closes one of the most common gaps in development processes — the disconnection between task tools and code tools. Previously, tickets often sat in “In Review” status with no clear visibility into who exactly was doing the review and for how long, and managers had to manually check GitHub or ping developers. By connecting review notifications to tickets, GitHub signals a strategic shift toward deep AI integration into project management, where the project tool and code repository behave as a unified system, and an AI agent participates in both layers simultaneously.