Skip to main content
Last updated on

Connect GitHub Copilot CLI

Use GitHub Copilot CLI with models, MCP tools, and skills through Unity Gateway. The Unity Gateway CLI (ug) handles workspace sign-in and configuration. You can also configure the connections manually.

Prerequisites​

You need your Databricks workspace URL and access to the models, MCP services, and skills you want to use. For manual setup, install GitHub Copilot CLI and have a Databricks personal access token.

If your admin has already configured your device, follow your organization's setup instructions.

Use the Unity Gateway CLI​

Install ug, then run this command from your project directory:

Bash
ug copilot

Follow the prompts to select your workspace and sign in. The command opens GitHub Copilot CLI with a configured model. Start working with the same prompts and commands you already use.

To start with a different model available in your workspace:

Bash
ug copilot --model "<model-name>"

Use the model's full Unity Catalog name, such as system.ai.claude-sonnet-4-6. The current ug integration configures Claude and GPT models for Copilot CLI.

Configure models manually​

Set these variables in your terminal:

Bash
export COPILOT_PROVIDER_TYPE="openai"
export COPILOT_PROVIDER_BASE_URL="https://<workspace-hostname>/ai-gateway/mlflow/v1"
export COPILOT_PROVIDER_BEARER_TOKEN="<databricks-pat>"
export COPILOT_MODEL="system.ai.claude-sonnet-4-6"
export COPILOT_OFFLINE="true"

Replace the hostname, token, and model name. The openai provider type selects the gateway's Chat Completions API, including when you use a Claude model. Offline mode disables connections to GitHub's servers; model requests still go to Unity Gateway.

Start Copilot CLI from the same terminal:

Bash
copilot

See GitHub's custom model provider documentation for these settings.

Add MCP tools​

To use system.ai.dbsql, system.ai.sandbox, or system.ai.web_search, an account admin must enable the Unity Gateway beta from the account console Previews page. See Manage account previews.

Use the Unity Gateway CLI​

Run:

Bash
ug mcp add --agents copilot

Select the services you want to add, then start a new session with ug copilot. Inside Copilot, enter /mcp to see connected servers. Ask the agent to use a tool by describing what you need.

ug copilot loads the MCP connections registered by ug. Keep using it to start your sessions.

Configure MCP services manually​

Merge this entry into ~/.copilot/mcp-config.json:

JSON
{
  "mcpServers": {
    "dbsql": {
      "type": "http",
      "url": "https://<workspace-hostname>/ai-gateway/mcp-services/system.ai.dbsql",
      "headers": {
        "Authorization": "Bearer <databricks-pat>"
      },
      "tools": ["*"]
    }
  }
}

Replace the hostname and token. For another service, use a unique server name and replace system.ai.dbsql with its three-part Unity Catalog name. Keep existing server entries, then restart Copilot CLI.

See GitHub's MCP documentation.

Add skills​

Use the Unity Gateway CLI​

Download published skills:

Bash
ug skills add

Choose the skills you want to use. Copilot CLI reads the downloads from ~/.agents/skills/. Start a new session, or enter /skills reload, then use /skills list to see them. Re-run the download command to get updated versions.

Configure skills manually​

For skill files you already have, place the complete folder, including SKILL.md and bundled files, in ~/.agents/skills/ or your project's .github/skills/. See GitHub's skills documentation.

To access the published skill registry through MCP, add a second server using the MCP configuration above. Name it databricks-skill-registry and use this URL with the same bearer authentication:

Text
https://<workspace-hostname>/ai-gateway/skills/

Keep the trailing slash. Restart Copilot CLI and ask it to use a published skill by its full name. Registry skills are accessed through MCP; /skills list shows installed skills.

Unity Gateway skills are in Beta. See Govern skills for enablement and permissions.

Troubleshooting​

A model request fails: Check the gateway URL, token expiration, full model name, and model permissions. The current ug integration excludes Gemini models from Copilot CLI because of a request compatibility issue.

An MCP server is missing: If you registered it with ug, restart with ug copilot. For manual setup, check ~/.copilot/mcp-config.json and the connection details in /mcp.

A downloaded skill is missing: Check that its folder contains SKILL.md in a supported directory, then enter /skills reload.

For workspace sign-in or ug setup problems, run ug doctor.

Next steps​

On this page