# Set up MCP

# Set Up deepset MCP Server

Configure the deepset MCP server and use it with your coding agent to interact with <ProductName /> pipelines.

## Prerequisites

- A <ProductShortName /> workspace and an API key. For safety, use a workspace-scoped API key. For instructions, see [Generate an API Key](/docs/how-to-guides/managing-access/generate-api-key.mdx).
- `uv` installed. If you don't have it, run:
  - Mac/Linux: `curl -LsSf https://astral.sh/uv/install.sh | sh`
  - Windows: `powershell -c "irm https://astral.sh/uv/install.ps1 | more"`
  - With Python already installed: `pipx install uv`

## Set Up the MCP Server

### In Cursor

Create a file named `mcp.json` in the `.cursor` directory at your project root:

```json
{
  "mcpServers": {
    "deepset": {
      "command": "uvx",
      "args": ["deepset-mcp"],
      "env": {
        "DEEPSET_WORKSPACE": "your-workspace-name",
        "DEEPSET_API_KEY": "your-api-key"
      }
    }
  }
}
```

`DEEPSET_WORKSPACE` is optional. If you leave it out, your agent must specify the workspace name when calling tools.
You can set it to the name of the workspace that contains the pipelines you want to investigate.

### Claude Desktop

Edit your Claude Desktop configuration file at `/Users/your-user/Library/Application Support/Claude/claude_desktop_config.json` (Mac):

```json
{
  "mcpServers": {
    "deepset": {
      "command": "uvx",
      "args": ["deepset-mcp"],
      "env": {
        "DEEPSET_WORKSPACE": "your-workspace-name",
        "DEEPSET_API_KEY": "your-api-key"
      }
    }
  }
}
```

`DEEPSET_WORKSPACE` is optional. If you leave it out, your agent must specify the workspace name when calling tools.
You can set it to the name of the workspace that contains the pipelines you want to investigate.