Copilot Memory Now Enabled by Default for Pro and Pro+ Users
GitHub CopilotGitHub has switched Copilot Memory from an opt-in feature to the default experience for all Copilot Pro and Pro+ subscribers. The system allows Copilot agents to automatically discover and store repository-specific facts — such as coding conventions, architectural patterns, and dependency relationships — and reuse that knowledge across future interactions in the coding agent, code review, and CLI. Memories are scoped to individual repositories, validated against the current codebase before use, and expire automatically after 28 days.
Sources & Mentions
4 external resources covering this update
Building an agentic memory system for GitHub Copilot
Hacker News
Let us git rid of it, angry GitHub users say of forced Copilot features
Hacker News
GitHub Copilot Evolves: SDK Launch, Agentic Memory & New AI Models (February 2026 Update)
Dev.to
Suggestion: Enable persistent memory in GitHub Copilot for better long-term collaboration
GitHub
Copilot Memory Is Now On by Default for Pro and Pro+ Subscribers
GitHub has made Copilot Memory the default experience for all Copilot Pro and Copilot Pro+ individual subscribers, transitioning the feature out of its opt-in phase within the existing public preview. Developers no longer need to navigate to settings to activate memory — it is now active automatically, allowing Copilot agents to build persistent, repository-level context starting from the first interaction.
What Copilot Memory Does
Copilot Memory enables Copilot agents to discover and store tightly scoped facts about a repository as they work. These include coding conventions and style preferences, architectural patterns and structural decisions, and critical cross-file dependencies that affect how changes ripple through the codebase. Rather than requiring developers to re-establish context at the start of every session or task, Copilot can draw on stored memories to deliver more relevant, consistent assistance.
The memory system operates with a key set of constraints designed to keep it accurate and safe. Every memory is strictly scoped to a single repository — no cross-repository bleed. Before a stored memory is applied, it is validated against the current state of the codebase to ensure it remains accurate; stale or outdated memories are not surfaced. All memories automatically expire after 28 days, preventing the accumulation of outdated assumptions.
Cross-Agent Memory Sharing
One of the more significant aspects of Copilot Memory is that it is shared across Copilot's agentic surfaces. Knowledge discovered during a coding agent session — such as a convention the agent encounters while implementing a pull request — is immediately available to the Copilot CLI and code review agents in the same repository. This cross-agent sharing means the system improves continuously across all development workflows without requiring developers to repeat themselves in each context.
User and Admin Controls
For individual users on Copilot Pro and Pro+, Copilot Memory is now active by default with no action required. Developers who prefer not to use the feature can disable it at any time from their personal Copilot settings under Features > Copilot Memory.
For organizations and enterprises, the feature remains under administrator control. Organization and enterprise administrators manage Copilot Memory availability through the standard Copilot policies settings, meaning enterprises can roll out or restrict the feature according to their own governance requirements.
Repository owners have granular control over stored memories: all memories for a given repository can be reviewed and deleted under Repository Settings > Copilot > Memory. This gives codebase owners a clear audit path and the ability to reset Copilot's accumulated context if needed.