# AdCP - Ad Context Protocol

> AdCP is an open standard published by AgenticAdvertising.org that enables AI agents to participate in advertising workflows. Built on Model Context Protocol (MCP) and Agent-to-Agent Protocol (A2A), AdCP defines a structured way for AI agents representing brands, agencies, publishers, and ad platforms to communicate and transact programmatically.

## Overview

AdCP (Ad Context Protocol) is the first open protocol designed specifically for agentic advertising. It provides a standard set of tasks, schemas, and transport mechanisms that allow AI agents to discover media products, negotiate and execute media buys, manage creative assets, enforce brand safety, and report on campaign performance — all without human-mediated API integrations.

The protocol is maintained by AgenticAdvertising.org, an open member organization of brands, agencies, publishers, ad tech platforms, and AI companies working to standardize how AI agents interact in the advertising ecosystem.

## Key concepts

AdCP is organized into eight protocol domains:

- **Media buy**: Product discovery, media buy creation, audience syncing, creative assignment, delivery reporting, and conversion tracking
- **Creative**: Canonical creative capability discovery, asset building, preview rendering, and delivery verification
- **Accounts**: Account listing, syncing, financial reporting, and usage tracking across platforms
- **Signals**: Audience and contextual signal discovery, activation, and data provider integration
- **Governance**: Content standards, property authorization (adagents.json), creative provenance, and delivery validation
- **Sponsored intelligence**: Conversational ad experiences where AI agents serve relevant brand information within chat interfaces
- **Brand protocol**: Machine-readable brand identity via brand.json files hosted at well-known URLs
- **Registry**: Centralized discovery of AdCP-compatible agents, brands, and publishers

## Transport protocols

AdCP supports two transport mechanisms:
- **MCP (Model Context Protocol)**: For tool-calling integrations where an AI agent invokes AdCP tasks as tools
- **A2A (Agent-to-Agent Protocol)**: For agent-to-agent communication where autonomous agents collaborate on advertising workflows

## Documentation

- Introduction: https://adcontextprotocol.org/docs/intro
- Getting started: https://adcontextprotocol.org/docs/building
- MCP integration guide: https://adcontextprotocol.org/docs/building/by-layer/L0/mcp-guide
- A2A integration guide: https://adcontextprotocol.org/docs/building/by-layer/L0/a2a-guide
- Media buy domain: https://adcontextprotocol.org/docs/media-buy
- Creative domain: https://adcontextprotocol.org/docs/creative
- Signals domain: https://adcontextprotocol.org/docs/signals/overview
- Governance domain: https://adcontextprotocol.org/docs/governance/overview
- Sponsored intelligence: https://adcontextprotocol.org/docs/sponsored-intelligence/overview
- Brand protocol: https://adcontextprotocol.org/docs/brand-protocol
- Schemas and SDKs: https://adcontextprotocol.org/docs/building/by-layer/L4/choose-your-sdk
- Release notes: https://adcontextprotocol.org/docs/reference/release-notes

## Adoption path

Build in this order:
1. Quickstart (5-minute curl example): https://adcontextprotocol.org/docs/quickstart
2. Build an agent: https://adcontextprotocol.org/docs/building/by-layer/L4/build-an-agent
3. Validate an agent: https://adcontextprotocol.org/docs/building/verification/validate-your-agent
4. Operate an agent: https://adcontextprotocol.org/docs/building/operating/operating-an-agent

Coding-agent shortcut — skip (2) and load a seller skill directly: https://github.com/adcontextprotocol/adcp-client/blob/main/skills/build-seller-agent/SKILL.md

## SDKs and skills

Each AdCP SDK ships both caller APIs (for buyers) and server primitives (for sellers). The JavaScript SDK also ships a storyboard runner (`npx @adcp/sdk@latest storyboard run`) that validates any endpoint against the spec — use it regardless of the language your agent is written in.

- `@adcp/sdk` (JavaScript/TypeScript): https://github.com/adcontextprotocol/adcp-client
- `adcp` (Python): https://github.com/adcontextprotocol/adcp-client-python
- `adcp-go` (Go): https://github.com/adcontextprotocol/adcp-go

Each SDK ships `build-*-agent` skills a coding agent can load to scaffold an agent and wire it to the storyboard runner. See https://adcontextprotocol.org/llms-full.txt for the full skill inventory.

## Registry APIs

Served by AgenticAdvertising.org:
- Agent directory: https://agenticadvertising.org/api/registry/agents
- Brand resolution: https://agenticadvertising.org/api/brands/resolve
- Property resolution: https://agenticadvertising.org/api/properties/resolve

## Key terms

- **Agentic advertising**: The practice of using AI agents to automate advertising workflows
- **Ad Context Protocol (AdCP)**: The open protocol specification
- **AgenticAdvertising.org**: The member organization that maintains AdCP
- **brand.json**: A machine-readable brand identity file hosted at /.well-known/brand.json
- **adagents.json**: A property authorization file hosted at /.well-known/adagents.json that declares which ad agents are authorized to serve ads on a domain
- **Task**: A discrete unit of work in AdCP (e.g., get_products, create_media_buy)
- **Media product**: An advertising opportunity offered by a publisher or platform
- **Content standards**: Brand safety and suitability rules that govern where ads can appear
- **Canonical format declaration**: A portable creative contract with `format_kind` and typed `params`, embedded in publisher catalogs, product `format_options[]`, and creative-agent capabilities.
- **capability_id**: An agent-local identifier from `get_adcp_capabilities.creative.supported_formats[]`, selected with `target_capability_id` when asking that creative agent to build.
- **format_id**: Deprecated in AdCP 3.2. Structured named-format references remain only for 3.x compatibility.

## Full documentation

For comprehensive protocol details, see: https://adcontextprotocol.org/llms-full.txt