> ## Documentation Index
> Fetch the complete documentation index at: https://docs.soriaanalytics.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Quick Start

> Connect Soria to Claude Code via MCP and run your first pipeline commands.

The fastest way to use Soria is through the MCP server, which lets Claude Code call all Soria tools directly. Follow these steps to connect and run your first commands.

<Steps>
  <Step title="Add the MCP proxy to Claude Code">
    The MCP proxy is a stdio server that connects Claude Code to any Soria backend. Add it to your global Claude Code config at `~/.claude.json`:

    ```json theme={null}
    {
      "mcpServers": {
        "soria": {
          "command": "uv",
          "args": ["--directory", "/path/to/soria", "run", "python", "-m", "soria.mcp_proxy"]
        }
      }
    }
    ```

    Replace `/path/to/soria` with the absolute path to your Soria repository. Alternatively, the project's `.mcp.json` (already in the repo root) configures this automatically when you open the project in Claude Code.
  </Step>

  <Step title="Start the backend">
    The MCP server runs as part of the Soria backend. Start it with:

    ```bash theme={null}
    make run-dev
    ```

    The backend starts on `http://localhost:8900`. The MCP endpoint is at `http://localhost:8900/mcp/`.
  </Step>

  <Step title="Discover available environments">
    Once connected in Claude Code, ask it to call `environment_list` to see available backends:

    ```
    Use environment_list to show me available Soria environments
    ```

    This returns all reachable environments — your local server, any deployed instances, and PR preview environments.
  </Step>

  <Step title="Switch to your environment">
    Use `environment_switch` to connect to a specific backend without restarting Claude Code:

    ```
    Use environment_switch to connect to local
    ```

    You can also switch to a deployed environment by name (e.g., `soria-streaming-extraction`) or a full URL.
  </Step>

  <Step title="Try your first commands">
    Explore what's available:

    ```
    List all dashboard pages
    ```

    ```
    Search for Medicare Advantage enrollment data
    ```

    ```
    Show me the available data scrapers
    ```

    To start building a pipeline, create a workspace first — all write operations require one:

    ```
    Create a workspace for a new scraper called my_data_source
    ```
  </Step>
</Steps>

## Direct HTTP connection (alternative)

If you prefer to connect directly to a specific backend without the proxy:

```bash theme={null}
claude mcp add --transport http soria http://localhost:8900/mcp/
```

Replace the URL with your deployment URL to connect to a remote instance. Note that direct connections don't support environment switching — you'll need to reconnect to change backends.

## What's next

<CardGroup cols={2}>
  <Card title="MCP Overview" icon="robot" href="/mcp/overview">
    Browse all available MCP tools by category
  </Card>

  <Card title="Pipeline Overview" icon="pipe-section" href="/pipeline/overview">
    Learn how the data ingestion pipeline works
  </Card>

  <Card title="Workspaces" icon="layer-group" href="/workspaces/overview">
    Understand workspace isolation before making data changes
  </Card>

  <Card title="Dashboards" icon="chart-bar" href="/dashboards/overview">
    Explore and query your data
  </Card>
</CardGroup>