Imported from abhi-singhs/copilot-cli-claude-code-compat (skills/claude-compat/SKILL.md). Install upstream with npx skills add abhi-singhs/copilot-cli-claude-code-compat --skill claude-compat. Copyright stays with the author.

Claude Code → Copilot CLI Reference

Use this reference when you know a Claude Code command and want the Copilot CLI equivalent.

CLI Subcommand Mapping

CLI Flag Mapping

Direct Match (no translation needed)

Translated (different name or syntax)

Unsupported (no Copilot equivalent)

Copilot CLI Only (no Claude Code equivalent)

Slash Command Mapping

Direct Matches (same in both CLIs)

/add-dir, /clear (/new, /reset), /compact, /context, /copy, /diff, /exit, /feedback, /help, /ide, /init, /login, /logout, /mcp, /memory, /model, /plan, /plugins (/plugin), /rename, /resume (/continue), /security-review, /skills, /tasks, /terminal-setup, /theme, /undo (/rewind), /usage, /quit

Renamed Commands

Claude Code Only (no Copilot equivalent)

/advisor [model|off], /autocompact [auto|<tokens>], /autofix-pr, /background (/bg), /chrome, /claude-api [migrate|managed-agents-onboard|prompt-audit], /color, /dataviz, /design-login, /design-sync, /desktop, /doctor (/checkup), /effort, /fast, /fewer-permission-prompts, /focus, /heapdump, /hooks, /import [codex|gemini] [--dry-run] [--yes], /list-agents (/peers), /loop (/proactive), /radio, /recap, /run, /run-skill-generator, /verify, /schedule (/routines), /scroll-speed, /setup-bedrock, /stats, /stop, /team-onboarding, /tui, /web-setup, /install-github-app, /powerup, /status, /teleport, /workflows, /reload-plugins

Copilot CLI Only (not in Claude Code)

/after [DELAY PROMPT] / /every [INTERVAL PROMPT] (experimental: schedule a one-shot or recurring prompt, skill, or schedulable slash command for this session, e.g. /after 30m remind me or /every 1h run tests; no args opens the schedule manager), /app (launch the GitHub Copilot app, or show the download URL if it isn't installed), /ask (experimental), /autopilot [OBJECTIVE] (/goal, experimental), /changelog (/release-notes) [summarize] [VERSION|last N|since VERSION], /chronicle (experimental: standup|tips|improve|reindex|skills create|skills review|skills status — session history tools, insights, and repository skill proposals), /clikit [COMPONENT] (internal/debug: preview CLI business components), /downgrade <VERSION> (download and restart into a specific CLI version; team accounts only), /env, /extensions (/extension) [manage|mode] (manage CLI extensions), /fleet, /list-dirs, /cwd (/cd), /lsp, /research, /restart, /refine [TEXT] (rewrite a roughly composed prompt into a clear one for review before sending), /rubber-duck [PROMPT] (consult the rubber duck agent for a second opinion on plans, code, and tests), /settings (/config) [--repo|--local] [show KEY|KEY|KEY VALUE] (open the settings dialog, focus it on a key, set a setting inline, or show a key's current value), /limits and /limits set max-ai-credits VALUE / /limits unset [max-ai-credits|all] (per-response AI credit limits), /move [branch|task] (experimental: move uncommitted changes into a new Git worktree and switch to it), /pr [view|create|fix|auto|automerge] (manage pull requests for the current branch), /reset-allowed-tools (revoke permissions granted during the session), /instructions (view the custom instruction files loaded into the current session), /sidekicks (view running sidekick agents), /streamer-mode (/on-air), /subagents (/agents) (configure default and per-agent subagent models), /user, /search [QUERY] (/find [QUERY]) (experimental: search the conversation timeline), /session (/sessions) with subcommands: info|checkpoints [n]|files|plan|rename [NAME]|cleanup|prune|delete [ID]|delete-all, /statusline (/footer), /experimental, /remote [on|off], /keep-alive [on|off|busy|DURATION] (/caffeinate), /tuikit [colors|icons|select|tabbar] (internal/debug: preview TUIkit design-system components and color tokens), /update (/upgrade), /version, /worktree [branch|task] and /worktree new [PROMPT] (experimental; see the note below)

Note: Copilot CLI has a /worktree slash command (experimental) in addition to the --worktree (-w) launch flag. /worktree [branch|task] creates a new Git worktree and switches the current conversation to it, leaving uncommitted changes behind in the current worktree — pass a branch name, a task description (used as the opening prompt), or nothing to auto-generate a branch name from the conversation. /worktree new [PROMPT] instead starts a new conversation in a new worktree and leaves the current conversation and its working directory unchanged; new is a reserved subcommand keyword, so it can't be used as a literal branch name. /move [branch|task] is a third variant that moves uncommitted changes into the new worktree and switches to it, where /worktree leaves them behind. All three require a Git repository, are experimental, and branch off HEAD unless the worktreeBaseRef setting is "defaultBranch". Claude Code has no exact equivalent: /fork (copy the conversation to a background session) and /branch are the closest analogs, but neither creates a worktree.

Note: Copilot CLI's --worktree (-w) launch flag (experimental) creates or reuses an isolated Git worktree under <repo>.worktrees/ at startup — matching Claude Code's --worktree (-w) flag. cpc maps claude --worktree NAME → copilot --worktree=NAME. Because Copilot's flag is experimental, enable experimental mode for it to take effect.

Note: /delegate is the Copilot equivalent of Claude Code's --cloud "task" flag. --remote is a deprecated alias for --cloud.

Note: /on-air (/streamer-mode) toggles streamer mode in Copilot CLI (hides preview model names and quota details for streaming). No Claude Code equivalent.

Note: /background (/bg) detaches the current session to run as a background agent. No direct Copilot equivalent; closest is Ctrl+X then b to promote to background.

Note: Copilot CLI's Agent and task delegation tools are task (delegate work to a background agent), list_agents (list running background agents), read_agent (read a background agent's output), and write_agent (send a message to a running background agent). Claude Code manages background agents through the claude CLI (attach, logs, respawn, stop) rather than these tools, so there is no direct cpc mapping — the workflows differ.

Note: /goal [condition|clear] sets a goal so Claude keeps working across turns until the condition is met. Copilot CLI's /autopilot [OBJECTIVE] (alias /goal, experimental) starts or refocuses autopilot mode with an optional objective — without one it infers intent from context — and supports --max-ai-credits N to cap spend; /goal on//goal off toggle autopilot without changing the objective, matching Claude Code's /goal <condition> → /goal off (for clear/stop/off/cancel).

Note: /stop stops the current background session (only available while attached). No Copilot CLI equivalent.

Note: /scroll-speed adjusts mouse wheel scroll speed interactively. No Copilot CLI equivalent.

Note: /theme options changed to [default|github|dim|high-contrast|colorblind].

Note: /code-review (Claude Code v2.1.x) replaces /simplify; /simplify and /review remain as backward-compatible aliases. All three map to Copilot CLI's /review. The --fix flag (apply the suggested fixes), the --comment flag (post inline PR comments), and effort levels (low|medium|high|xhigh|max|ultra) have no Copilot equivalent. Copilot CLI's /review [PROMPT] takes a free-form prompt, so those arguments must be stripped instead of passed through — cpc does this automatically when /review ... is used as the initial prompt (e.g. cpc -p "/review high --fix pr#123" → copilot -p "/review pr#123"), but in-session slash commands can't be translated.

Note: /ultraplan was removed from Claude Code. Use plan mode (/plan) instead, which exists in both CLIs.

Note: /usage-credits is the renamed /extra-usage (Claude Code v2.1.x): "configure usage credits to keep working when you hit a limit". No Copilot CLI equivalent — the closest is /usage which only shows usage stats.

Note: /run, /run-skill-generator, and /verify (Claude Code v2.1.145+) are skills that build, launch, and drive the project's app to observe a change running. No Copilot CLI equivalent.

Note: /radio opens Claude FM lo-fi radio in the browser (not available on Bedrock, Vertex, or Foundry). No Copilot CLI equivalent.

Note: /dataviz [request] (Claude Code v2.1.198+) is a skill that provides design guidance for charts, graphs, and dashboards — Claude picks the chart form, assigns color by role, and validates the palette for colorblind safety. No Copilot CLI equivalent.

Note: /design-login authorizes design-system access for /design-sync using your claude.ai account (requires a claude.ai subscription). No Copilot CLI equivalent.

Note: /design-sync [hint] is a skill that converts your repo's React design system and uploads it to Claude Design so generated designs use real components (a first-time sync can take a few hours on a large repo). Only available on the Anthropic API — not on Amazon Bedrock, Google Cloud Vertex AI, or Microsoft Foundry. No Copilot CLI equivalent.

Note: /compact [FOCUS-INSTRUCTIONS] now accepts optional focus instructions in both CLIs (e.g. /compact focus on the auth module). The cpc wrapper doesn't touch in-session slash commands, so the focus argument passes through to Copilot CLI unchanged.

Note: /autocompact [auto|<tokens>] (Claude Code v2.1.221+) sets how full the context window gets before Claude Code compacts automatically, and saves the value to user settings. Copilot CLI has no equivalent — its /compact is a one-shot compaction command, not a threshold setting. The --autocompact <auto|tokens> launch flag (same setting, session-only) is likewise unsupported and dropped by cpc with a warning.

Note: /claude-api [migrate|managed-agents-onboard|prompt-audit] is a Claude Code-only skill. The prompt-audit subcommand (v2.1.221+) flags instructions written for older models in prompts, skills, and tool descriptions and proposes fixes as a diff. No Copilot CLI equivalent, and no impact on the cpc wrapper.

Note: /sandbox exists in both CLIs but with different syntax. Claude Code's /sandbox toggles sandbox mode; Copilot CLI's /sandbox [config|status|policy|enable|disable] manages OS-level sandboxing that restricts filesystem and network access for shell commands, MCP/LSP servers, and built-in file/web tools. config (or bare /sandbox) opens the sandbox settings dialog, status reports whether sandboxing is enabled, policy shows the effective policy including path grants, denials, and network access, and enable/disable turn it on or off directly. The subcommands are additive, so /sandbox enable and /sandbox disable still work as before. Copilot CLI additionally has --sandbox / --no-sandbox launch flags (experimental mode only) that enable or disable the OS-level shell sandbox for a single session without changing the saved setting — Claude Code has no launch-flag equivalent.

Note: /permissions differs between the CLIs. Claude Code manages persistent allow/ask/deny tool rules with /permissions. Copilot CLI's /permissions [default|assisted|allow-all|show] now switches between permission modes (this is new — it previously only supported show/reset); /permissions reset remains available as a separately documented subcommand to clear in-memory tool and path approvals. /allow-all (/yolo) is documented as an alias for /permissions allow-all, and its on option was replaced with auto (/allow-all [off|auto|show]).

Note: /rubber-duck [PROMPT] is a Copilot CLI-only agent for a second opinion on plans, code, and tests. No Claude Code equivalent.

Note: /bug [report] is now the primary bug-reporting command in Claude Code, with /share as its alias; /feedback still exists and opens the same dialog. Before v2.1.212, /bug and /share were both aliases of /feedback. In the VS Code extension, /bug opens the extension's own feedback dialog (v2.1.229+). The /share alias collides in name with Copilot CLI's /share [link|off|file|html|gist|research] [...] (session sharing and export). Same name, different action — /share submits a bug report in Claude Code but shares or exports the session in Copilot CLI.

Note: /context [all] accepts an optional all argument in Claude Code, which expands the per-item context breakdown in fullscreen rendering (previously the breakdown always collapsed so the grid stayed visible). Copilot CLI's /context takes no arguments, so drop all when switching.

Note: /doctor is also available as /checkup in Claude Code. Both check the health of the installation and auto-updater. No Copilot CLI equivalent.

Note: /import [codex|gemini] [--dry-run] [--yes] (Claude Code v2.1.213+) brings configuration from another coding agent already on the machine into Claude Code — instruction files, MCP servers, commands, subagents, and skills. It currently supports OpenAI Codex and Google Gemini CLI, and is not available on Amazon Bedrock, Google Cloud's Agent Platform, Microsoft Foundry, or Claude Platform on AWS. --dry-run previews the changes without writing; --yes skips the interactive picker. There is no Copilot CLI equivalent and nothing for cpc to map — the command imports into Claude Code, and Copilot CLI is not one of the supported sources.

Note: /list-agents (alias /peers, Claude Code v2.1.224+) lists the subagents and other Claude Code sessions Claude can message, along with the name to use for each. It's only available when cross-session messaging is enabled. Copilot CLI has no direct equivalent: /subagents (alias /agents) configures subagent models rather than listing peers, and the list_agents tool only lists background agents started from the current session, not other sessions.

Note: /clear [name] in Claude Code accepts an optional name to label the previous conversation in the /resume picker. Copilot CLI's /clear [PROMPT] instead takes an optional prompt to start the new conversation. The optional argument has different semantics on each CLI.

Note: /deep-research <question> (Claude Code workflow: "fan out web searches on a question, fetch and cross-check sources, and synthesize a cited report") has no exact Copilot CLI equivalent. Closest is /research [TOPIC], which uses GitHub search + web sources rather than fanned-out web search. The cpc wrapper treats this as a best-effort translation.

Note: /fork changed semantics twice. It was originally an alias for /branch; in v2.1.161 it became a forked-subagent command; and in v2.1.212 it changed again — /fork [prompt] now copies the current conversation into a new background session while the original session keeps running, and the two sessions are independent after the copy. That matches Copilot CLI's own /fork [NAME] (experimental; added v1.0.45), which forks the current session into a new independent session, so the two CLIs are now aligned. The forked-subagent behavior moved to /subtask (see below). On v2.1.161–v2.1.211, /fork <directive> still spawns a forked subagent.

Note: /subtask <task> (Claude Code v2.1.212+) spawns a forked subagent that inherits the full conversation and works on the task while you keep working in the main session; its result returns to the conversation when the subagent finishes. Copilot CLI has no 1:1 equivalent — the closest is /fleet <task> (parallel subagent execution), so cpc treats it as a best-effort mapping. Use /fork in Claude Code to copy the conversation into a separate background session instead.

Note: /btw [question] (Claude Code) asks a side question without adding it to the conversation history. The question argument became optional in v2.1.212 — running /btw with no argument reopens the overlay on the session's most recent side question. Copilot CLI's /ask (experimental) is the equivalent.

Note: /advisor [model|off] enables or disables the Claude Code server-side advisor tool interactively (accepts opus, sonnet, or a full model ID; no argument opens a picker; requires v2.1.98+). The fable alias was removed in v2.1.212 and is no longer accepted. No Copilot CLI equivalent.

Note: /cd <path> moves the current session to a new working directory (Claude Code v2.1.169+; preserves the prompt cache and appends the new directory's CLAUDE.md as a message). Copilot CLI's /cd [PATH] (combined with /cwd) covers the same action.

Note: /reload-skills (Claude Code v2.1.152+) re-scans skill and command directories so skills added or changed on disk become available without restarting. Maps to Copilot CLI's /skills reload.

Note: /config (Claude Code v2.1.181+) accepts inline key=value pairs (e.g. /config thinking=false) to set a setting directly without opening the Settings interface; the key=value form also works in non-interactive mode (-p) and from Remote Control. Copilot CLI now documents /config as an alias of /settings, so the command name matches — use a space instead of = (/config thinking false).

Note: /effort [low|medium|high|xhigh|max|ultracode|auto] gained the ultracode level in Claude Code v2.1.203 (combines xhigh reasoning with automatic workflow orchestration; like max, it's a session-only option). Copilot CLI has no ultracode level — cpc maps the --effort/--reasoning-effort ultracode flag to max with a warning.

Note: /settings and /config are the same command in Copilot CLI — /config [--repo|--local] [show KEY|KEY|KEY VALUE] opens the settings dialog, opens it focused on a specific KEY, sets a setting inline with a KEY VALUE pair, or prints a key's current value with show KEY (secret-named values are masked). --repo and --local target .github/copilot/settings.json and .github/copilot/settings.local.json instead of user settings, and only repo-overridable keys can be set that way. Because Copilot CLI now documents /config as an alias, Claude Code's /config is a name match — only the inline syntax differs (/config thinking=false in Claude Code vs /config thinking false in Copilot CLI).

Note: /security-review [PROMPT] is now a direct match: both Claude Code and Copilot CLI run a security review agent that analyzes pending changes for vulnerabilities. The cpc wrapper passes in-session slash commands through unchanged.

Note: /subagents (alias /agents) configures the default and per-agent subagent models in Copilot CLI. It's a richer counterpart to Claude Code's /agents than /agent (which only browses agents).

Note: /after [DELAY PROMPT] and /every [INTERVAL PROMPT] are Copilot CLI-only experimental commands that schedule a one-shot or recurring prompt, skill, or schedulable slash command for the current session (e.g. /after 1h /chronicle standup, /every 1d run tests); no arguments opens the schedule manager. Claude Code's /schedule (/routines) only partially overlaps — there is no exact equivalent.

Note: /extensions (alias /extension) [manage|mode] manages Copilot CLI extensions. Claude Code's /plugin is conceptually similar but refers to a distinct plugin system, so there is no exact mapping.

Note: /app launches the GitHub Copilot desktop app (or shows the download URL if it isn't installed). No Claude Code equivalent.

Note: /mcp gained a search subcommand in Copilot CLI (/mcp [show|add|edit|delete|disable|enable|auth|reload|search] [SERVER-NAME]) for searching available MCP servers.

Note: /refine TEXT is a Copilot CLI-only command that rewrites a roughly composed prompt into a clear one so you can review it before sending. Run it with no arguments (via Ctrl+X then /refine) to clean up whatever is already in the input box — particularly useful for prompts entered by speaking. Claude Code has no equivalent, and because it is an in-session slash command the cpc wrapper can't translate it.

Note: /voice now exists in both CLIs with different arguments. Claude Code's /voice [hold|tap|off] selects the input style (push-to-talk, tap, or off); Copilot CLI's /voice [on|off|models|devices] toggles voice mode, browses available voice models, or chooses the input device (microphone). on/off are the only overlapping arguments, so drop hold/tap when switching. Both are in-session slash commands, so cpc doesn't translate them.

Keyboard Shortcuts

Global

Timeline

Navigation

Session Picker (opened via /resume or --continue)

Environment Variables

Copilot CLI Only

Supported Models

The --model / /model value is passed straight through in both CLIs. Copilot CLI currently supports:

Config Directory Mapping

Tool Permission Syntax

Copilot CLI's --allow-tool / --deny-tool patterns take the form Kind(argument), and the argument is optional. The kinds are a fixed set — memory, read, shell, url, write, or an MCP server name — not tool names, so several Claude Code tools collapse onto the same kind. (Claude Code makes a similar simplification: since v2.1.210 it warns that Glob(path) and NotebookEdit(path) rules should be written as Read(path) or Edit(path).)

For shell rules, the :* suffix matches the command stem followed by a space — shell(git:*) matches git push and git pull but not gitea. Deny rules always take precedence over allow rules, even with --allow-all. There is no glob or grep permission kind; those are tool names, used only with --available-tools / --excluded-tools.

Tool Availability Names (--tools → --available-tools)

Copilot CLI also exposes list_bash/read_bash/stop_bash/write_bash (and their PowerShell twins), apply_patch, list_agents, read_agent, and write_agent, which have no Claude Code launch-flag counterpart.