Get started with coding agents

Unity Gateway connects coding agents such as Claude Code, Codex, and OpenCode to models from multiple providers, MCP tools, and shared skills. Choose a model for the task, give your agent access to company data, and use your team's workflows.

You can connect in two ways:

  • Use ug (recommended): Install the Unity Gateway CLI, sign in, and launch your preferred agent. ug handles the agent-specific connection settings and applies your organization's configuration, so you can start coding without editing config files or writing setup scripts.
  • Set up the connection yourself: Enter the gateway endpoint and authentication settings in your agent. See Supported coding agents to choose your agent and open its guide for desktop and CLI setup.

This guide walks you through using ug.

Note

If your admin has already installed ug and configured your device, go straight to launching your agent.

Before you begin

You need:

  • Your Azure Databricks workspace URL and access to the models you want to use. Ask your admin if you don't have these.
  • Python 3.12 or later and uv.

1. Install ug

In your terminal, run:

uv tool install git+https://github.com/databricks/unity-gateway

To check the installation, run:

ug --version

You should see the installed version number.

2. Launch your agent

Open a terminal in your project directory. Run the command for the agent you want to use:

Claude Code

ug claude

Codex

ug codex

Gemini CLI

ug gemini

OpenCode

ug opencode

GitHub Copilot CLI

ug copilot

Pi

ug pi

On first use, follow the prompts to select your workspace and sign in. ug handles the connection to Unity Gateway, then opens your chosen agent with the same interface and workflow you already know. You can start using it right away.

To switch models during a session, use /model to choose from those available through your organization.

3. Add tools and skills

In your terminal, run:

ug mcp add
ug skills add

Choose the MCP servers and skills you want to add from those you have access to. MCP tools connect your agent to data and services; skills provide instructions for specific tasks.

Then launch or restart your agent to use what you added:

Command inside your agent What you can do
/mcp View the MCP servers you added and their connection status.
/skills Browse the skills you added and select one for your task.

Slash commands vary by agent. Check its help menu for model, MCP, and skill commands. See Add tools and skills for more setup options.

4. Check your usage

In your terminal, run:

ug usage

You'll see how much you've spent and how much of your budget you've used.

Next steps

  • Last updated on