GitHub makes Copilot Memory generally available on paid plans

GitHub made Copilot Memory generally available on paid plans.

GitHub made Copilot Memory generally available on paid Copilot plans on October 28. Memory lets Copilot use personal preferences and repository conventions across sessions. The announcement lists the Copilot app, CLI, VS Code, cloud agent, code review, Slack, Teams and GitHub Mobile as places where memories can be created. GitHub also announced public preview for dreaming, a distinct feature that revisits past work to curate memories or propose guidance changes. [GitHub changelog](https://github.blog/changelog/2026-10-28-github-intelligence-puts-your-code-context-and-history-to-work/)

A saved convention can help a later session

A personal memory might record how a developer wants a pull request description organized. A repository memory might record which test is relevant for a particular behavior. Those details can be useful when a developer starts a new session from another client and would otherwise need to repeat the same explanation. The distinction between personal and repository context also matters when the guidance concerns a team convention and should not be inferred from one person’s habits.

GitHub says people can view or delete saved preferences and describes settings for repository access and dreaming. That matters because repository practices change. An instruction that was correct before a migration can become stale, and a useful memory feature needs a path to find and revise it. GitHub’s public-preview observations describe sessions it reviewed in which memory was used. They should not be treated as a general productivity prediction. [GitHub changelog](https://github.blog/changelog/2026-10-28-github-intelligence-puts-your-code-context-and-history-to-work/)

Dreaming has a separate rollout and policy

The preview covers personal memory curation, repository customization and code review customization. GitHub says organization and enterprise use is governed by Copilot policies, with personal dreaming also requiring the user’s own setting. The GA label for Memory does not carry over to this separate preview. [GitHub changelog](https://github.blog/changelog/2026-10-28-github-intelligence-puts-your-code-context-and-history-to-work/)

A team can evaluate Memory by saving a non-sensitive repository convention, verifying that it appears in a fresh session and then editing or deleting it. That checks whether the feature helps a real workflow and whether the owner can correct what the agent carries forward.