GAL CLI

Governance platform for AI coding agents. Discover, centralize, and sync approved configurations for individuals and teams.

Install

pnpm add -g @scheduler-systems/gal-run

Or via Homebrew:

brew tap scheduler-systems/tap
brew install gal

Windows:

winget install SchedulerSystems.GAL

Bootstrap scripts:

curl -fsSL https://gal.run/install.sh | sh

powershell -NoProfile -ExecutionPolicy Bypass -Command "$script = Join-Path $env:TEMP 'gal-install.ps1'; Invoke-WebRequest -UseBasicParsing -Uri 'https://gal.run/install.ps1' -OutFile $script; & $script; Remove-Item $script -Force"

Usage

# Authenticate with GitHub
gal auth login

# Pull approved configs
gal sync --pull

# Configure MCP + validate tool discovery
gal mcp setup
gal mcp doctor

# Check sync status
gal sync --status

# Scan org repos for AI agent configs
gal discover

Internal Preview

gal code is currently an internal-only command gated behind the existing internal audience tier. It does a live GET /auth/me check before launch, so authenticated non-internal users are rejected even if they guess the command. The production request path is gal code -> GAL API proxy -> Vertex, not direct client access to Vertex.

Current local preview flow:

cd /Users/scheduler-systems/Documents/scheduler-systems/products/gal-run-private
pnpm build:gal-code
pnpm build:cli
GAL_NO_AUTO_UPDATE=1 \
GAL_ACCEPT_TERMS=true \
node apps/cli/dist/index.cjs code --version

Manual smoke test:

cd /Users/scheduler-systems/Documents/scheduler-systems/products/gal-run-private
GAL_NO_AUTO_UPDATE=1 \
GAL_ACCEPT_TERMS=true \
node apps/cli/dist/index.cjs code

Expected behavior:

• startup lands in GAL Code with Model gal
• /model shows Current model gal
• non-internal authenticated users get GAL Code requires membership in an org with internal audienceTier
• prompt traffic goes to https://api.gal.run/api/gal-code/v1/chat/completions

Supported Platforms

• Claude Code (.claude/)
• Cursor (.cursor/)
• GitHub Copilot (.github/)
• Gemini (.gemini/)
• Codex (.codex/)
• Windsurf (.windsurf/)

MCP Server

GAL provides an MCP server for AI coding agents. Auto-configured by gal sync --pull, or set up directly:

gal mcp setup
gal mcp doctor

Recommended onboarding:

gal auth login
gal mcp setup
gal mcp doctor

Manual config example:

Claude Code (.mcp.json):

{
  "mcpServers": {
    "gal": {
      "type": "http",
      "transport": "http",
      "url": "https://api.gal.run/mcp"
    }
  }
}

Cursor (.cursor/mcp.json), Windsurf (.windsurf/mcp_config.json), Gemini (settings.json):

{
  "mcpServers": {
    "gal": {
      "type": "http",
      "transport": "http",
      "url": "https://api.gal.run/mcp"
    }
  }
}

The hosted endpoint https://api.gal.run/mcp uses OAuth/Bearer auth.

Codex Setup

Preferred (OAuth):

codex mcp add gal --url https://api.gal.run/mcp
codex mcp login gal

If OAuth login fails with Dynamic client registration not supported, use bearer-token mode:

export GAL_AUTH_TOKEN="$(node -e 'console.log(JSON.parse(require(\"fs\").readFileSync(process.env.HOME+\"/.gal/config.json\",\"utf8\")).authToken||\"\")')"
codex mcp remove gal
codex mcp add gal --url https://api.gal.run/mcp --bearer-token-env-var GAL_AUTH_TOKEN

If /mcp shows Tools: (none), verify GAL_AUTH_TOKEN is set in the Codex process environment.

CI/CD Integration

- name: Install GAL CLI
  run: pnpm add -g @scheduler-systems/gal-run

- name: Sync configs
  run: gal sync --pull
  env:
    GAL_TOKEN: ${{ secrets.GAL_TOKEN }}

Documentation

• Docs: https://gal.run/docs
• Dashboard: https://app.gal.run
• Issues: https://github.com/Scheduler-Systems/gal-run/issues

License

Proprietary - Copyright (c) 2026 Scheduler Systems. All rights reserved.