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

# A2: Your first media buy

> Module A2: Execute your first AdCP media buy. Free 20-minute hands-on module walking through discovery, purchase, creative sync, and delivery with a live sandbox agent.

# A2: Your first media buy

> [!NOTE]
>
> **Free module** — No account required. \~20 minutes with Addie. Prerequisite: [A1](https://docs.adcontextprotocol.org/dist/docs/3.3.0-rc.0/learning/foundations/a1-agentic-advertising.md).

## Learning objectives

* Execute the full media buy lifecycle: discovery, purchase, creative sync, delivery
* Identify the agent roles involved: buyer agent, sales agent, creative agent, signals agent
* Read and understand real protocol messages at each stage
* Recognize that 3.1 products may describe creative requirements with canonical `format_options[]` that narrow shared `format_kind` contracts
* Observe a live agent-to-agent transaction
* Read a capability declaration: some objects signal support by being present, while others carry an explicit `supported` flag that can be `false`
* Recognize the two things that make agent-to-agent requests safe: the request is signed (so the seller can verify who sent it) and it carries an `idempotency_key` (so a retry doesn't double-book)

## Reading list

> [card] [Protocol architecture](https://docs.adcontextprotocol.org/dist/docs/3.3.0-rc.0/protocol/architecture.md)
>
> The complete architecture: domain map, identity layer, transaction domains, governance, and ecosystem layers.

> [card] [MCP integration guide](https://docs.adcontextprotocol.org/dist/docs/3.3.0-rc.0/building/by-layer/L0/mcp-guide.md)
>
> How MCP works in practice: tool calls, response format, context management, async operations.

> [card] [Capability discovery](https://docs.adcontextprotocol.org/dist/docs/3.3.0-rc.0/protocol/get_adcp_capabilities.md)
>
> How agents advertise their capabilities so other agents can discover what they offer.

> [card] [AdAgents.json](https://docs.adcontextprotocol.org/dist/docs/3.3.0-rc.0/governance/property/adagents.md)
>
> Publisher authorization for which agents can sell or enrich inventory.

> [card] [Seller setup](https://docs.adcontextprotocol.org/dist/docs/3.3.0-rc.0/brand-protocol/seller-setup.md)
>
> How sellers use brand.json with adagents.json for identity and supply path verification.

> [card] [A2A guide](https://docs.adcontextprotocol.org/dist/docs/3.3.0-rc.0/building/by-layer/L0/a2a-guide.md)
>
> The Agent-to-Agent protocol — how specialized agents collaborate on complex campaigns.

> [card] [Task lifecycle](https://docs.adcontextprotocol.org/dist/docs/3.3.0-rc.0/building/by-layer/L3/task-lifecycle.md)
>
> How tasks move through states: from request to completion, including async operations.

## Key terms

| Term | Definition |
| - | - |
| **Sales agent** | Represents a publisher and exposes inventory via `get_products`. May also implement the Creative Protocol to handle creatives from the same endpoint. |
| **Buyer agent** | Represents a brand or agency and purchases media via `create_media_buy` |
| **Brand agent** | Manages brand identity and guidelines via `brand.json` |
| **Creative agent** | Any agent implementing the Creative Protocol — produces and adapts advertising assets via `build_creative`. This can be a standalone service or a sales agent that declares `"creative"` in `supported_protocols`. |
| **Signals agent** | Provides measurement and audience data via `get_signals` |
| **brand.json** | Public identity record for an advertiser, publisher, seller, or platform: name, brand assets, agents, and signing-key discovery |
| **adagents.json** | Publisher-hosted file declaring properties and the agents authorized to sell or enrich them |
| **Tool discovery** | The process of an agent reading another agent's capabilities to know what tasks it supports |
| **Request signing** | A cryptographic signature on the request that lets the seller verify who sent it and that it wasn't tampered with in transit |
| **Idempotency key** | A unique tag on each mutating request that lets a buyer safely retry without creating a duplicate |
| **Domain** | A broad protocol area an agent supports — `media_buy`, `creative`, `signals`, `governance`, `brand`, `sponsored_intelligence`. Declared via `supported_protocols`. |
| **Specialism** | A narrow capability claim within a domain — e.g. `sales-guaranteed`, `creative-generative`, `signal-marketplace`. Declared via `specialisms`. See the [Compliance Catalog](https://docs.adcontextprotocol.org/dist/docs/3.3.0-rc.0/building/verification/compliance-catalog.md). |
| **Canonical formats** | AdCP 3.1's v2 path for creative requirements: products can publish `format_options[]` that narrow shared creative archetypes such as `image` or `video_hosted`. Validate against the canonical first, then the product's narrowing. |
| **Storyboard** | A scripted compliance scenario shipped by the protocol at `/compliance/{version}/` — agents don't write them, they're run against agents. An agent demonstrates its domain and specialism claims by passing the matching storyboards. |

## Protocol versioning

Every request carries `adcp_major_version`; sellers advertise supported versions on `get_adcp_capabilities`. A3 and B1 go deeper on version negotiation.

## Reading capabilities

`get_adcp_capabilities` also tells buyers what a seller supports, and each capability follows its own schema contract. Optional feature objects such as `media_buy.audience_targeting` signal support by being present. Other blocks carry an explicit flag, and some must be declared even when the answer is no: every agent declares `adcp.idempotency`, with `{"supported": false}` when it does not deduplicate retried requests. Read each capability the way its schema defines it rather than applying one rule to all of them. B1 goes deeper on how sellers declare capabilities.

## What you'll do with Addie

Tell Addie what you want: audience, goals, budget. Then walk through each step as it happens:

1. **Discovery** — `get_products` against `@cptestagent`, examine real response structure
2. **Purchase** — `create_media_buy` with targeting and budget. Addie points out two things on this request: it carries a signature (so the seller can verify the buyer) and an `idempotency_key` (so a retry after a network error never creates two buys). Note the `confirmed_at` timestamp in the response — this is the seller's order confirmation.
3. **Creative** — `sync_creatives` to deliver assets to the publisher
4. **Status check** — `get_media_buys` to see lifecycle state, creative approvals, and `valid_actions`
5. **Delivery** — `get_media_buy_delivery` to see results

You'll see the actual protocol messages at each stage. By the end, you've bought media through an agent.

## Assessment

| Dimension | Weight | What Addie looks for |
| - | - | - |
| Conceptual understanding | 25% | Can you describe the transaction flow and which agent handles what? |
| Practical knowledge | 35% | Can you direct a media buy and interpret the delivery report? |
| Problem solving | 15% | Can you reason about what happens when things go wrong? |
| Protocol fluency | 25% | Do you use correct task names and agent roles? |

Passing threshold: 70%.

## Start this module

> [card] [Start A2 with Addie](https://agenticadvertising.org/chat)
>
> Open Addie and say "I'd like to start certification module A2."

**Next:** [A3: The AdCP landscape](https://docs.adcontextprotocol.org/dist/docs/3.3.0-rc.0/learning/foundations/a3-ecosystem-governance.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.