Migrate from GitHub Copilot to Augment Code.
2 documentation-derived translation patterns — what carries over and what to watch for. Cited to the Feature Parity Map; the audit tells you whether the move is worth it.
Augment Chat and Copilot Chat (Ask mode) cover the same job: an in-editor conversational assistant that explains code, plans changes, and answers questions grounded in your codebase, with one-click apply of suggested blocks. A team on Copilot can retire Augment Chat: use the Copilot Chat panel (or inline chat) in VS Code / Visual Studio / JetBrains and pick 'Ask' from the mode dropdown. Translate Augment's 'Smart Apply' to Copilot's 'Apply in Editor' / Insert action on a code block. Reach for codebase context the Copilot way — add #codebase or @workspace and #file/#selection references in the prompt — to replace what Augment's Context Engine pulled in automatically. Keep Copilot; drop Augment Chat.
- Warning: Grounding is not automatic in Copilot the way it is in Augment: Augment Chat answers from the Context Engine's whole-repo index by default, whereas Copilot Ask mode prioritizes open files and explicit #codebase/@workspace/#file references — expect to scope context manually or answers may miss distant code.
- Warning: Copilot Ask mode is intentionally lighter-touch than its Agent mode: it suggests and you apply, so a multi-file change that Augment Chat could stage may need Copilot's Edit or Agent mode instead of Ask.
- Warning: Chat quota and slash commands differ by plan and surface: the Copilot Free tier caps chat (~50 messages/month) and the exact slash commands (/explain, /fix, /tests, /new, /doc, etc.) vary per IDE — confirm the team's paid plan covers expected chat volume before cancelling Augment.
Augment Agent and Copilot Agent mode are the same workflow: an in-IDE autonomous agent that breaks a request into a plan, edits/creates/deletes code across the workspace, runs terminal commands behind approval, and uses MCP tools. A team already on Copilot can drop the standalone Augment seat: in VS Code (or Visual Studio / JetBrains) open Copilot Chat and pick 'Agent' from the Ask/Edit/Agent dropdown, then give it the same natural-language tasks. Map the approval habit directly — Augment's default 'Agent' (pause before terminal/tool calls) corresponds to Copilot's per-command confirmation, and 'Agent Auto' corresponds to enabling Copilot's auto-approve/allow-list for terminal and tools. Re-point any Augment MCP servers at Copilot's MCP config (mcp.json / the tools picker) so the agent keeps the same database/Linear/Jira reach. Keep Copilot; cancel Augment.
- Warning: Augment's edge is its Context Engine (semantic index of the whole repo); Copilot agent mode reads the workspace and searches on demand rather than from a prebuilt index, so on very large monorepos prime it explicitly with #codebase, #file, or @workspace references instead of assuming it already knows distant files.
- Warning: Billing model differs: Augment meters pooled team 'credits', while Copilot agent usage as of 2026-06-01 is metered in GitHub AI Credits by the selected model's token consumption (legacy annual cohort still on the premium-request multiplier) — re-estimate run-rate before cutting over, don't assume parity.
- Warning: No project carries over automatically: there is nothing to export from Augment Agent itself, but re-create any Augment rules/guidelines as Copilot custom instructions (.github/copilot-instructions.md) so the agent keeps your conventions.