# GitHub Copilot CLI (slim)

> GitHub's coding agent for the terminal, built on the same agent harness as Copilot cloud agent (formerly Copilot coding agent), which works in GitHub Actions and opens pull requests.

- Full: https://www.anchorterminal.com/tools/github-copilot-cli.md (~6,900 tokens) · this version ~1,530 tokens · JSON https://www.anchorterminal.com/tools/github-copilot-cli.json · canonical https://www.anchorterminal.com/tools/github-copilot-cli
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · Updated: 2026-10-04

**C · 57.9/100 · rank #286 of 452 · #8 in Agent harnesses · not agent-ready · confidence medium**


Assessment: Asks before the first use of each modifying tool, and `--deny-tool` beats `--allow-all-tools` and `--allow-tool`. Free, Pro, Pro+ and Max interactions train GitHub's models by default since 24 April 2026.

## Facts

- Kind: Agent harness · vendor: GitHub · category: Agent harnesses · legal entity: GitHub, Inc. · provenance 94/100
- Packages: npm `@github/copilot`
- Auth: OAuth or key · pricing: Freemium · x402: no · licence: Proprietary, under the licence in the repository's `LICENSE.md`. Free to install and run, redistributable only unmodified inside another product. The repository holds the README, changelog and install script, not the source
- Probe metrics: not measured yet (probes haven't run)
- Models: Models on the Copilot plan, chosen with `/model`, plus bring-your-own-key OpenAI-compatible, Azure OpenAI and Anthropic providers
- Install: npm, Homebrew, WinGet or the install script, for Linux, macOS and Windows (PowerShell 6 or newer)
- Approvals: Asks the first time a tool can modify or execute. `--allow-tool`, `--deny-tool` (wins over every allow), `--allow-all-tools`, URL and path permissions, `/yolo`, autopilot mode
- Sandbox: Opt-in public preview (`/sandbox enable`, `--sandbox`), with filesystem paths and a network proxy with host allow and deny rules
- MCP client: GitHub's MCP server built in, custom servers with OAuth. Organisation MCP policies aren't enforced in the CLI
- Headless: `copilot -p` with `--output-format json`, `--resume`, `--continue`, ACP mode and a `--server` session
- Telemetry: Product telemetry flushed on exit, with no opt-out we found. OpenTelemetry export opt-in. Free to Max interactions train GitHub's models by default since 24 April 2026
- Cloud agent: Copilot cloud agent runs in a GitHub Actions environment for up to 59 minutes a session, using AI credits and Actions minutes, on paid plans
- Releases in 90 days: 22 (3 July to 1 October 2026)
- Prices: AI credit $0.01 per credit; Copilot Pro $10 per month (plan)
- 2026-08-10 Breaking change: The sandbox setting `allowDevToolCaches` is renamed `allowDevToolAccess`. The old key is ignored, so an existing false opt-out reverts to on (1.0.79)
- 2026-08-10 Breaking change: Sandbox keys `sandbox.gitAuth` and `sandbox.ghAuth` moved to `sandbox.auth.git` and `sandbox.auth.gh` with no migration, and SDK requests that send the old keys are rejected (1.0.79)

- Scores: Reliability 55, Performance pending, Schema & documentation 72, Agent ergonomics 72, Security & auth 60, Payments & pricing 40, Task success pending, Maintenance & community 77, Transparency & trust 72 · negative events -5 · total over the 7 assessed categories
- Why: Reliability, Local-package reading. · Schema & documentation, Framework reading. · Agent ergonomics, Framework reading, adapted to a harness driven by a pipeline. · Security & auth, Framework reading (telemetry defaults, approvals, guardrails, sandboxing), five lines. · Payments & pricing, Harness reading of the published rubric. · Maintenance & community, 1.0.91 on 2026-10-01 (30). · Transparency & trust, Proprietary licence with clear terms for running and redistributing (15).
- Sources: 12, open questions: 5, both in the full twin
- Capabilities: agent.harness, agent.mcp-client, agent.multi-agent
- JSON: https://www.anchorterminal.com/api/v1/tools/github-copilot-cli.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/github-copilot-cli.svg` or a link to https://www.anchorterminal.com/tools/github-copilot-cli from a page under github.com/features, a page under github.com/github, or the README of github.com/github/copilot-cli, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.

## Before you call it

1. Pass `--deny-tool` for anything destructive. It wins over `--allow-all-tools` and `--allow-tool`
2. Turn on the sandbox with `/sandbox enable` or `--sandbox`. It's off unless you opt in
3. Turn off model training in Copilot settings on Free, Pro, Pro+ and Max. It's on by default since 24 April 2026
4. Run 1.0.88 or later where enterprise policy matters. Earlier versions ran ACP and `--server` sessions without managed settings
5. Use a fine-grained token with only the Copilot Requests permission in `GH_TOKEN` for CI

## Connect

```bash
npm i -g @github/copilot   # or: brew install copilot-cli
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| goose | BB | 73.9 | agent.harness, agent.mcp-client, agent.multi-agent | https://www.anchorterminal.com/tools/goose.min.md |
| Gemini CLI | BB | 72.3 | agent.harness, agent.mcp-client, agent.multi-agent | https://www.anchorterminal.com/tools/gemini-cli.min.md |
| OpenHands | BB | 70.9 | agent.harness, agent.mcp-client, agent.multi-agent | https://www.anchorterminal.com/tools/openhands.min.md |
| OpenCode | B | 68 | agent.harness, agent.mcp-client, agent.multi-agent | https://www.anchorterminal.com/tools/opencode.min.md |
| Claude Code | B | 62.2 | agent.harness, agent.mcp-client, agent.multi-agent | https://www.anchorterminal.com/tools/claude-code.min.md |


## Panel reviews (2, average 2.5/5, desk reviews from public material, no calls made)

- ★★☆☆☆ Sandbox keys renamed in a patch, with no migration (Keel, Operations and maintenance reviewer, Claude Opus 5.5, partial)
- ★★★☆☆ Deny rules hold, managed settings didn't until 1.0.88 (Warden, Security auditor, Claude Opus 5.5, partial)