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

# What is MCP?

> Introduction to the Model Context Protocol and how it works with Trackyard

The **Model Context Protocol (MCP)** is an open standard that lets AI agents like Claude and Cursor access external tools and data sources. Trackyard's MCP server gives AI agents direct access to the music catalog.

<Info>
  **Pricing Note:** MCP access may have different pricing than direct API access. Check [app.trackyard.com/api-keys](https://app.trackyard.com/api-keys) for current pricing details.
</Info>

***

## How It Works

Instead of calling the Trackyard API manually, you describe what you need in natural language to your AI agent:

**You:** "Find me upbeat music for a 30-second Instagram Reel about travel"

**Claude (via MCP):**

1. Searches Trackyard's catalog
2. Presents top results
3. Downloads the track when you confirm
4. Trims it to exactly 30 seconds

All in one conversation. Zero code.

***

## Why Use MCP Instead of the REST API?

<Tabs>
  <Tab title="MCP Server (Conversational)">
    **Best for:** Non-technical users, quick one-off searches, exploratory workflows

    **How it works:**

    * Talk to Claude/Cursor in plain English
    * The AI calls the Trackyard API for you
    * You confirm selections and the AI downloads tracks

    **Example workflow:**

    ```
    You: "I need music for a 15-second product video"
    AI: "What kind of vibe?"
    You: "Upbeat and modern, no vocals"
    AI: [searches, presents 3 options]
    You: "Download the second one"
    AI: [downloads and saves the file]
    ```
  </Tab>

  <Tab title="REST API (Programmatic)">
    **Best for:** Automation, pipelines, batch processing, custom integrations

    **How it works:**

    * Write code to call Trackyard endpoints
    * Integrate into your existing workflows
    * Automate hundreds/thousands of requests

    **Example workflow:**

    ```python theme={null}
    tracks = search("upbeat music")
    audio = download(tracks[0]["id"], duration_seconds=15)
    save("product_video_music.mp3", audio)
    ```
  </Tab>
</Tabs>

***

## Supported AI Agents

<CardGroup cols={2}>
  <Card title="Claude (Desktop)" icon="message">
    Anthropic's Claude desktop app with MCP support built-in.
  </Card>

  <Card title="Cursor" icon="code">
    AI-powered code editor with MCP integration.
  </Card>

  <Card title="OpenClaw" icon="robot">
    AI assistant platform with native MCP support.
  </Card>

  <Card title="Custom Agents" icon="gear">
    Any agent that implements the MCP standard.
  </Card>
</CardGroup>

***

## What You Can Do With MCP

<AccordionGroup>
  <Accordion title="Search by natural language" icon="magnifying-glass">
    **Example:**

    "Find me chill lo-fi music for a coding stream"

    **What happens:** The AI searches Trackyard, interprets your intent, and presents relevant tracks.
  </Accordion>

  <Accordion title="Download with smart trimming" icon="scissors">
    **Example:**

    "Download that second track as a 30-second clip"

    **What happens:** The AI calls `/download-track` with `duration_seconds: 30` and saves the file locally.
  </Accordion>

  <Accordion title="Align hit points" icon="bullseye">
    **Example:**

    "Trim it to 22 seconds with the big hit at 12 seconds"

    **What happens:** The AI uses `hit_point_seconds: 12` to align the musical peak.
  </Accordion>

  <Accordion title="Batch operations" icon="layer-group">
    **Example:**

    "Find 5 different tracks for travel videos and download 15-second clips of each"

    **What happens:** The AI searches once, then downloads 5 custom clips.
  </Accordion>
</AccordionGroup>

***

## When to Use MCP vs. API

**Use MCP when:**

* You want conversational music discovery
* You're doing one-off searches (not automation)
* You prefer natural language over code
* You're using Claude, Cursor, or another MCP-compatible agent

**Use the REST API when:**

* You're building automation or pipelines
* You need to process hundreds/thousands of requests
* You're integrating into your own application
* You want full programmatic control

***

## Next Steps

<CardGroup cols={2}>
  <Card title="Setup Guide" icon="gear" href="/mcp-server/setup-guide">
    Install the Trackyard MCP server
  </Card>

  <Card title="Using with AI Agents" icon="robot" href="/mcp-server/using-with-ai-agents">
    Learn how to use Trackyard via MCP
  </Card>

  <Card title="API Quickstart" icon="rocket" href="/getting-started/api-quickstart">
    Or use the REST API directly
  </Card>
</CardGroup>