GitHub Copilot Memory is generally available, with "dreaming" and cross-repository code search

GitHub introduces "GitHub intelligence", a context layer for Copilot agents built from GitHub data, with three parts:

• Memory: generally available on paid Copilot plans. Copilot remembers personal preferences and repository conventions across sessions and clients (Copilot app, CLI, VS Code, cloud agent, code review, Slack, Teams, GitHub Mobile). GitHub says that among Copilot CLI sessions reviewed in the preview where the agent used memory, 93% showed better adherence to developer preferences and 42% more efficient workflows.
• Dreaming: in public preview and opt-in. Copilot reflects on past work to curate personal memories, manage repository memories and propose updates to skills and custom instruction files (as pull requests for review), and turn recurring accepted code-review feedback into guidance for code review. Dream runs consume AI credits; administrators enable it through Copilot policies.
• Code search: Copilot can search code, configuration, schemas and documentation across repositories in an organization without a local clone, by exact text, regular expression or natural-language description, for example through /research.

Availability and settings vary by capability.