> ## Documentation Index
> Fetch the complete documentation index at: https://agenticadvertisingorg-snap-format-preview-links.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Getting Started

> **New to AdCP?** Start with the [**Quickstart Guide**](/dist/docs/2.5.3/quickstart) for a complete introduction including authentication, testing, and your first request.

## Quick Overview

AdCP works with multiple protocols, but you only need to learn one. Both protocols provide access to the same advertising tasks - choose based on your technical needs.

## Which Protocol Should I Use?

Both protocols provide access to ALL AdCP capabilities - the same tasks, same features, same functionality. Choose based on your technical ecosystem:

### Use MCP If:

* **You're using Claude** or other MCP-compatible AI assistants
* **You have existing MCP tools** in your workflow
* **You prefer direct tool calling** patterns
* **Your client supports MCP** natively

**Technical style**: Tool-based, request/response pattern

→ [**Start with the MCP Guide**](/dist/docs/2.5.3/protocols/mcp-guide)

### Use A2A If:

* **You're using Google's agent ecosystem** or A2A-compatible agents
* **You prefer task-based** interactions
* **You like Server-Sent Events** for real-time updates
* **Your client supports A2A** natively

**Technical style**: Task-based, message/artifact pattern

→ [**Start with the A2A Guide**](/dist/docs/2.5.3/protocols/a2a-guide)

## What You'll Be Able to Do

Regardless of protocol choice, you can:

* Discover advertising products and inventory
* Create and manage media buys
* Upload and assign creative assets
* Monitor campaign delivery
* Manage audience signals
* Handle targeting and optimization

## How AdCP Protocols Work

```
Your Application
       ↓
[Choose One Protocol]
       ↓
   MCP or A2A
       ↓
  AdCP Tasks
       ↓
Ad Platforms (Google, Meta, Amazon, etc.)
```

Both protocols are just different ways to access the same underlying AdCP tasks. The business logic and capabilities are identical.

## Next Steps

1. **Choose your protocol** based on the criteria above
2. **Read the relevant guide** (MCP or A2A)
3. **Try the examples** in your chosen protocol
4. **Explore the tasks** available in AdCP

## Quick Comparison

Both protocols support all features - they just express them differently:

### MCP Example

```json theme={null}
// Request
{
  "method": "get_products",
  "params": {
    "brief": "CTV campaign for sports fans"
  }
}

// Response (flat structure - task fields at top level)
{
  "status": "completed",
  "message": "Found 5 CTV products",
  "context_id": "ctx-123",
  "products": [...],
  "clarification_needed": false
}
```

### A2A Example

```json theme={null}
// Request
{
  "message": {
    "parts": [{
      "kind": "text",
      "text": "CTV campaign for sports fans"
    }]
  }
}

// Response
{
  "contextId": "ctx-123",
  "artifacts": [{
    "artifactId": "artifact-product-catalog-xyz123",
    "name": "product_catalog",
    "parts": [
      {"kind": "text", "text": "Found 5 CTV products"},
      {"kind": "data", "data": {"products": [...]}}
    ]
  }]
}
```

Same data, different structure. Both can handle clarifications, async operations, approvals, and all AdCP features.

## Don't Get Overwhelmed

You don't need to read everything! Start with:

1. This page (you're here ✓)
2. Your chosen protocol guide
3. Try some examples

The reference documentation is there when you need it, but most users only need the protocol guides.

## Getting Help

* **For MCP questions**: See the [MCP Guide](/dist/docs/2.5.3/protocols/mcp-guide)
* **For A2A questions**: See the [A2A Guide](/dist/docs/2.5.3/protocols/a2a-guide)
* **For context & task IDs**: See [Context Management](/dist/docs/2.5.3/protocols/context-management) to understand the difference between `context_id` and `task_id`
* **For task details**: See the [Media Buy](/dist/docs/2.5.3/media-buy) or [Signals](/dist/docs/2.5.3/signals/overview) sections
* **For deep technical details**: Check the [Reference](/dist/docs/2.5.3/protocols/error-handling) section (only if needed)