◀ Playbook index
NO.19

Copilot Memory

Updated: 2026-10-26

In a nutshell

Copilot Memory is a mechanism for Code Review / Cloud Agent / CLI to remember and reuse knowledge learned in a repository.

No more repeating the same preamble every time — the more you use it, the smarter Copilot becomes.

Scope and mechanics

Scope

  • 📦 Per repository — memories tied to a repo, shared across users who have access
  • 👤 Per user (Pro / Pro+, early access) — personal preferences that follow you across all repos and Copilot agents, without affecting others. Examples: preferred commit style, PR structure, tone. Manage them in your personal Copilot Memory settings ↗
  • 🔀 Cross-feature — knowledge learned by Cloud Agent is available to Code Review, CLI, and vice versa

How it works

  • 🔍 Each memory carries code citations that are verified against the current codebase when the memory is used
  • ⏳ Auto-deleted after 28 days — refreshed and extended whenever the memory is used

Why it matters

No more repeating the same explanation in every prompt — the more you use it, the deeper Copilot’s understanding of the repository grows.

Where to configure it

ScopeSettings pathWhat you can do
RepositoryRepo → Settings → Code and automation → Copilot → MemoryView the list of saved memories and delete individual entries
OrganizationOrg → Settings → Code, planning, and automation → Copilot → Policies → Copilot MemoryEnable / Disable for the entire org
EnterpriseEnterprise → AI controls → Copilot → Copilot MemoryEnabled / Disabled / Let orgs decide for all orgs

📘 Official guide: Managing and curating Copilot Memory

Slides 日本語