Skip to content

github-copilot.md

Latest commit

 

History

History
103 lines (74 loc) · 3.62 KB

github-copilot.md

File metadata and controls

103 lines (74 loc) · 3.62 KB
Raw

Claude Code with GitHub Copilot as Model Provider

Guidance for how to connect GitHub Copilot as a model provider for Claude Code, using copilot-gateway (npm).

NOTICE: calling GitHub Copilot is not against its policy as this is officially supported per doc here. And actually, there are lots of AI tools (e.g. Aider and Cline VSCode extension) already support GitHub Copilot as one of the LLM providers.

1) Install Claude Code

npm install -g @anthropic-ai/claude-code

Copilot Gateway itself doesn't need a global install — run it directly with npx.

2) Start copilot-gateway and authenticate to GitHub Copilot

$ npx copilot-gateway@latest start --proxy-env
...
Please visit https://github.com/login/device and enter code XXXX-XXXX to authenticate
...

Once succeeds, the server listens on http://localhost:4141 and exposes both OpenAI (/v1/chat/completions, /v1/responses, /v1/models, /v1/embeddings) and Anthropic (/v1/messages) compatible endpoints. List available models with:

curl -s http://localhost:4141/v1/models | jq -r '.data[].id'

Useful start options (run npx copilot-gateway@latest start --help for the full list):

Option Description
--port, -p Port to listen on (default: 4141)
--account-type, -a Account type: individual, business, enterprise
--proxy-env Initialize proxy from environment variables
--claude-code, -c Generate a ready-to-paste Claude Code launch command
--rate-limit, -r Rate limit in seconds between requests
--api-key Require API keys for incoming requests

3) Create Claude Code configure file ~/.claude/settings.json with the following contents

{
  "env": {
    "ANTHROPIC_BASE_URL": "http://localhost:4141",
    "ANTHROPIC_AUTH_TOKEN": "sk-dummy",
    "ANTHROPIC_MODEL": "claude-sonnet-5",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "gpt-5-mini",
    "DISABLE_NON_ESSENTIAL_MODEL_CALLS": "1",
    "CLAUDE_CODE_ATTRIBUTION_HEADER": "0"
  }
}

Model names must match what your gateway actually serves — check with the /v1/models request above and replace as needed.

Alternatively, run with --claude-code to get a ready-to-paste launch command without editing settings:

npx copilot-gateway@latest start --claude-code

4) Run claude

Open another terminal and then run claude at your will. DO read its best practices for fully leveraging its capabilities.

Alternative config

If the above-configured file doesn't work, use the env variable directly:

export ANTHROPIC_BASE_URL="http://localhost:4141"
export ANTHROPIC_AUTH_TOKEN="sk-dummy"
export ANTHROPIC_MODEL="claude-sonnet-5"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="gpt-5-mini"
export DISABLE_NON_ESSENTIAL_MODEL_CALLS="1"
export CLAUDE_CODE_ATTRIBUTION_HEADER="0"

claude

Docker

Pre-built images are published to ghcr.io/feiskyer/copilot-gateway on every release:

docker run -p 4141:4141 ghcr.io/feiskyer/copilot-gateway:latest

# Persist auth across restarts with a bind mount
mkdir -p ./copilot-data
docker run -p 4141:4141 \
  -v $(pwd)/copilot-data:/root/.local/share/copilot-gateway \
  ghcr.io/feiskyer/copilot-gateway:latest

Usage monitoring

Check your Copilot usage/quota without starting the server:

npx copilot-gateway@latest check-usage