copilot

Copilot plugin for Claude Code

Use GitHub Copilot from inside Claude Code for code reviews, delegated tasks, model selection, and resumable background runs.

I liked OpenAI's codex-plugin-cc, and I use Claude Code, Codex, and Copilot regularly. I wanted the same kind of Copilot workflow inside Claude Code too, so I built this.

This project is an open-source Claude Code companion plugin inspired by openai/codex-plugin-cc, but wired to the GitHub Copilot CLI instead of Codex.

Why this exists

I did not want Copilot to be the odd one out in my workflow.

• Claude Code is where I like to work.
• Codex already had a good plugin story.
• I still wanted to call Copilot for reviews, rescue tasks, and model-specific runs without leaving CC.

This plugin is the bridge for that.

What you get

• /copilot:setup to check local Copilot CLI readiness and show the configured default model
• /copilot:review for a structured read-only review of your current git changes
• /copilot:adversarial-review for a more challenging review focused on risks and design tradeoffs
• /copilot:rescue to hand a coding task to Copilot, optionally with --model, --effort, or --reasoning-effort
• /copilot:status, /copilot:result, and /copilot:cancel for background-job management

Requirements

• GitHub Copilot CLI installed locally
• An active GitHub Copilot entitlement
• Node.js 18.18 or later

Install

Add the marketplace in Claude Code:

/plugin marketplace add codywilliamson/copilot-plugin-cc

Install the plugin:

/plugin install copilot@codywilliamson-copilot

Reload plugins:

/reload-plugins

Install Copilot CLI if needed:

npm install -g @github/copilot

Authenticate with either:

copilot login

or:

gh auth login

Then run:

/copilot:setup

Usage

/copilot:review

Review the current working tree, or compare your branch to a base ref.

/copilot:review
/copilot:review --base main
/copilot:review --model gpt-5.4
/copilot:review --background

/copilot:adversarial-review

Run a sterner review that actively questions assumptions, edge cases, and safer alternatives.

/copilot:adversarial-review
/copilot:adversarial-review --base main question the retry strategy and rollback plan
/copilot:adversarial-review --model claude-sonnet-4.6 --background focus on race conditions

/copilot:rescue

Delegate a coding task to Copilot. This command is write-capable by default and is the main “go do the work” path.

/copilot:rescue investigate why the tests started failing
/copilot:rescue --model gpt-5.4 --effort high fix the flaky integration test
/copilot:rescue --model gpt-5.4 --reasoning-effort xhigh plan a safe migration for this schema change
/copilot:rescue --resume apply the smallest safe fix from the last run
/copilot:rescue --background refactor this module into smaller pieces

Notes:

• --model <id> is passed through to Copilot CLI.
• --effort <low|medium|high|xhigh> and --reasoning-effort <low|medium|high|xhigh> are both accepted and forwarded to Copilot.
• If you omit --model, Copilot CLI uses its configured default model from ~/.copilot/config.json.

/copilot:status

Show current and recent Copilot jobs for this repository.

/copilot:status
/copilot:status task-abc123

/copilot:result

Show the stored final output for a completed or failed job.

/copilot:result
/copilot:result task-abc123

/copilot:cancel

Cancel an active background run.

/copilot:cancel
/copilot:cancel task-abc123

How it works

The plugin shells out to copilot -p in JSON output mode, captures Copilot’s session and result events, and stores per-repo job state under Claude Code’s plugin data directory. That gives Claude Code a thin, open wrapper around the local Copilot CLI without trying to replace Copilot itself.