GitHub Copilot for Jira: Public Preview Enhancements

GitHub Copilot for Jira Gets Meaningful Upgrades in Public Preview

GitHub has shipped a meaningful set of improvements to the Copilot coding agent's Jira integration, responding to early feedback from teams that adopted the feature since its public preview debut on March 5, 2026. The update addresses four distinct pain points: onboarding complexity, model control, work traceability, and context richness.

Model Selection Directly from Jira

Previously, the AI model powering the Copilot coding agent was determined by organization policy or default settings — there was no way to influence it from within Jira itself. The update changes that. Developers can now specify the model they want to use by including the model name in their comment when mentioning @GitHub Copilot on a Jira issue. This gives teams per-task control over the intelligence-vs-speed tradeoff without leaving the Jira interface.

Jira Ticket Numbers in Pull Requests

One of the most commonly requested improvements was better cross-tool traceability. The Copilot coding agent now automatically includes the originating Jira ticket number in the pull request title and branch name. The pull request also includes a back-link to the Jira issue and a snapshot of the context that was provided to the agent.

Confluence Integration via Atlassian MCP

The most significant new capability in this update is the ability to connect the Copilot coding agent to Confluence through the Atlassian MCP (Model Context Protocol) server. By configuring the MCP server with a personal access token (PAT), developers can grant the agent read access to Confluence pages — design documents, architecture specs, product requirements, and similar reference material that lives outside the codebase.

Improved Onboarding and Error Guidance

GitHub has also strengthened the onboarding path by providing clearer error messages when common configuration issues arise, and by expanding the integration's prerequisite documentation.