# Unity CLI as the replacement for the in-Editor MCP server

> Learn how the Unity CLI replaces the MCP server in the in-editor AI assistant package, what the deprecation affects, and which MCP tools remain available.

Unity has deprecated the [MCP server](https://docs.unity3d.com/Packages/com.unity.ai.assistant@2.18/manual/integration/unity-mcp-landing.html) included in the in-editor AI assistant package. The Unity CLI replaces this MCP server. This page explains the scope of the deprecation, which tools are unaffected, and how to move your agent to the Unity CLI.

## MCP server deprecation

Unity has deprecated the MCP server component of the in-editor AI assistant package `com.unity.ai.assistant`. This server runs in the Unity Editor, and your MCP client connects to it directly. The `unity mcp` command in the Unity CLI replaces the in-Editor server. The Unity CLI includes its own MCP server, built on the [Unity Pipeline package](/unity-production-pipeline/local-tools-cli/unity-pipeline-package.md). It uses the same protocol, so your MCP client connects the same way it does today.

The MCP mode in the Unity CLI is not part of the deprecation. It remains fully supported. Some agents and harnesses can't run arbitrary shell commands, and some run models that don't compose command lines reliably. MCP mode supports these cases.

### Direct commands for the Editor

The `unity command` and `unity eval` commands drive the Editor directly, without MCP in between. They're faster and use fewer tokens than MCP. Use these commands if your agent can run shell commands.

### Third-party MCP packages

Third-party MCP packages from GitHub that you installed to get agent access into Unity are unaffected.

### Check whether the deprecation affects you

| Your setup                                                                              | Affected |
| --------------------------------------------------------------------------------------- | -------- |
| The in-editor AI assistant package with its MCP server enabled to connect an MCP client | Yes      |
| A third-party MCP package installed from GitHub                                         | No       |
| The `unity mcp` or `unity command` commands from the Unity CLI                          | No       |

## Migrate to the Unity CLI

To move an existing MCP client to the CLI:

1. Install the Unity CLI.
2. Run `unity pipeline install`.
3. Configure your MCP client to use the Unity CLI's MCP server. Run `unity mcp configure <client>`. Replace `<client>` with the name of your client, for example, `claude`, `cursor`, `vscode`, or `windsurf`. This updates the client's configuration file to use the Unity CLI's MCP server instead of the in-Editor MCP server.
4. Run `unity mcp configure --list` to see all supported clients and their current configuration status.
5. Install the Unity skills for your agent. Run `unity skill install <agent-name>`.

After you complete these steps, your agent uses the Unity CLI, instead of the MCP server in the in-editor AI assistant package, to interact with the Unity Editor.

For more information, refer to [Install Unity CLI](/unity-cli/use-unity-cli.md#install-the-unity-cli).

### Unity CLI capabilities

Before you migrate to the Unity CLI, learn about its capabilities:

* **Editor versions**: The CLI manages any Editor version. To drive a running Editor through the Unity Pipeline package, use Unity 6.0 LTS or later.
* **Network access**: The CLI drives a local Editor offline. It runs on your machine, and the local pipeline server accepts connections on localhost only.
* **Licensing**: The CLI is free and separate from Unity AI. Installing the CLI doesn't require a Unity AI subscription. The CLI works with the agent you already use, or with no agent.
* **AI agent usage**: The CLI also works with no AI agent at all. Use it to:
  * Install and manage Editor versions.
  * Manage projects.
  * Run headless builds.
  * Run test suites.

These capabilities make the CLI usable as a continuous integration (CI) tool on its own.

> **Note:**
>
> Versions 2.13 and later of the in-editor AI assistant package fix a conflict between the CLI and the in-editor AI assistant package. If your project uses an earlier version, update the package.