Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Important
This feature is in Beta. To use it, a workspace admin must turn on Unity Gateway Managed Configuration from the Previews page. See Manage Azure Databricks previews.
Use a coding agent configuration to give your team a shared setup for agents, models, MCP servers, and skills. Publish the configuration for developers in your Azure Databricks workspace, and the Unity Gateway CLI (ug) fetches it and applies it on each developer's machine.
For configuration schema reference, see Coding agent configuration schema.
Coding agent configurations
A workspace admin publishes a coding agent configuration to a Azure Databricks workspace. Developers apply it to their machines with ug configure.
Note
Publishing a configuration requires a workspace admin. Developers can read and apply a published configuration with either the Workspace access or Consumer access entitlement. See Manage entitlements.
The configuration defines:
| Setting | What it configures |
|---|---|
| Coding agents | The enabled agents and the default agent. |
| Models | Each agent's model source and default models. Use Azure Databricks models or an external provider registered in Unity Catalog. |
| MCP servers | The servers to connect to the enabled agents. |
| Skills | The shared skills to make available to the enabled agents. |
| Unity Gateway features | Agent settings such as Smart Routing and tracing, where supported. |
Reference models, MCP servers, and skills by their Unity Catalog names, or select them from a catalog and schema. For external models, see Bring your own model providers.
Prepare resources
Before you publish, set up the models your team needs and grant developers access to them, along with any MCP servers and skills you include. See Set up model capacity.
Publish a configuration
Use the UI
As a workspace admin:
- Open Unity Gateway > Govern > Agent configuration.
- Click Edit configuration to open the JSON editor.
- Update the agents, models, MCP servers, skills, and features in the configuration. Click Format JSON to format the JSON.
- Click Save to publish the configuration for developers in this workspace.


Choose model sources for each agent
For each enabled agent, set sources in config.models to a list of the models to make available. Agents in the same configuration can use different lists. Every model must support an API that is compatible with the agent. For example, Claude Code requires the Anthropic Messages API. Codex requires the Open Responses API or OpenAI Responses API. GitHub Copilot supports the Anthropic Messages API, the OpenAI Responses API, and the Chat Completions API.
Each entry is a model service such as system.ai.claude-haiku-5-5, a model behind a model provider service such as main.default.anthropic_prod.claude-sonnet-5-5, or a wildcard. catalog.schema.* selects every model service in the schema, and catalog.schema.model-provider-service.* selects every model behind that model provider service. A schema wildcard requires at least one model service in the schema, and a provider wildcard requires the provider service to exist. A wildcard must be the entire last segment. The list accepts at most 100 entries, which must be non-blank and unique.
{
"models": {
"sources": [
"main.default.anthropic_prod.claude-sonnet-5-5",
"system.ai.claude-haiku-5-5",
"main.default.*",
"main.default.openai_prod.*"
]
}
}
Each default in default_models must be a specific model that the list includes, either listed directly or matched by a wildcard.
With Smart Routing on, the router selects among every model the list includes, including external models behind a model provider service. When two entries resolve to the same model, Smart Routing uses only the first entry in list order. To choose which model provider serves a model, place its entry before the entry for the other provider.
For every models field, see the configuration schema.
For Smart Routing model-source requirements, see Smart Routing limitations.
Configuration example
This example configures Claude Code with a default model, one MCP server, and one skill. Replace the resource names with resources your team can access. When editing an existing configuration, preserve any settings your team still needs.
For every supported field, see Coding agent configuration schema.
{
"default_agent": "CODING_AGENT_CLAUDE_CODE",
"enabled_agents": [
{
"agent": "CODING_AGENT_CLAUDE_CODE",
"config": {
"models": {
"sources": ["system.ai.claude-sonnet-5-5"]
},
"default_models": {
"default_model": "system.ai.claude-sonnet-5-5"
}
}
}
],
"mcp_servers": { "names": ["main.developer_tools.github"] },
"skills": { "names": ["main.team_skills.code_review"] }
}
The default_agent must also be listed in enabled_agents.
For details about the overall default and Claude Code family defaults, see Set a default model.
Configure GitHub Copilot
One CODING_AGENT_GITHUB_COPILOT entry configures both Copilot CLI and Copilot in Visual Studio Code. It requires default_models.default_model and vscode_integration.enabled:
true:ug configuresets up Copilot CLI and Copilot in Visual Studio Code when Visual Studio Code is installed. Developers can skip Visual Studio Code with--no-vscode.false:ug configuresets up only Copilot CLI and removes an earlier Visual Studio Code setup.
Smart Routing isn't supported for GitHub Copilot. For an example entry, see Coding agent configuration schema. Developers must run the latest version of ug to apply GitHub Copilot settings. Ask them to run ug upgrade.
Use the API
To publish from a script, save the JSON as coding-agent-config.json. With the Azure Databricks CLI authenticated to the workspace as a workspace admin, create the configuration:
databricks api post /api/ai-gateway/v2/coding-agent-configs \
--json @coding-agent-config.json
The workspace supports one published configuration. To change it, update the existing configuration with PATCH /api/ai-gateway/v2/coding-agent-configs/{id}, using the returned resource name and an update_mask query parameter that identifies the fields to change. Keep the JSON file in version control to review changes before publishing them.
Sync configuration updates
When developers launch an agent through ug, it checks the published configuration and applies updated agent and model settings. Changes can take a few minutes to appear.
To refresh immediately and sync managed MCP servers and skills, run:
ug configure
For GitHub Copilot in Visual Studio Code, developers restart Visual Studio Code after ug configure.
Enable developers to view spend
Configure shared or per-user budget thresholds with Block usage enabled. Developers can then run ug usage to view their applicable spend limits, or ug usage --all to view every limit.
For setup instructions, see Manage budgets for Unity Gateway. For command details, see Check your usage.