Skip to main content
AdCP 3.0 BetaYou’re viewing the v3 beta documentation. The API is stable for testing, but breaking changes may occur before final release. See what’s new in 3.0 and the v3 migration guide. For production use, switch to version 2.5 using the version selector above.

Introduction to AdCP

Welcome to the Ad Context Protocol (AdCP) documentation—the open standard for agentic advertising.

The Opportunity

RTB unlocked programmatic. AdCP unlocks the rest. 90% of ad spend never touches RTB—it lives in walled gardens, direct deals, and premium inventory. Execution costs limit how many media partners advertisers can work with. The opportunity isn’t optimizing existing platforms better—it’s expanding to more partners without scaling headcount. AI agents collapse this complexity cost. AdCP gives them a standard way to buy media, build creatives, and activate audiences across any platform.

What is AdCP?

AdCP is an open standard for advertising automation that enables AI agents to interact with advertising platforms through unified interfaces:
  • One Protocol: Single interface for all advertising platforms
  • AI-Native: Works over MCP and A2A protocols for seamless agent integration
  • Platform Agnostic: Works with any compatible advertising platform
AdCP uses a task-first architecture where core advertising tasks (like creating media buys or discovering signals) can be accessed through multiple protocols:
  • MCP (Model Context Protocol): For direct AI assistant integration
  • A2A (Agent2Agent Protocol): For complex workflows and agent collaboration
All tasks and data models are defined with JSON schemas for validation and client code generation.

Protocol Architecture

AdCP operates at multiple layers, providing a clean separation between the business roles, orchestration layer, and technical execution:

The AdCP Ecosystem Layers

Top Layer: Business Principals

Buying Principal (Left) The demand side of advertising, including:
  • Advertisers: Brands with products/services to promote
  • Agencies: Acting on behalf of advertisers
  • Retail Media Networks: Retailers monetizing their audiences
  • Curators: Packaging inventory and data for specific use cases
Media Seller (Right) The supply side of advertising, including:
  • Publishers: Content creators with audience reach
  • Sales Houses: Representing multiple publishers
  • Rep Firms: Specialized sales representation
  • SSPs: Supply-side platforms aggregating inventory
  • Ad Networks: Aggregating and reselling inventory
These parties exchange impressions and money through the orchestration layer below.

Middle Layer: Orchestration

Media Orchestration Platform (Left) Platforms that evaluate sellers and audiences, and execute buying strategies:
  • Examples: Scope3, custom orchestration solutions
  • Function: Strategy execution, seller evaluation, optimization
  • Integration: Uses MCP to communicate with both Audience and Sales Agents
Signal Agent (Right, Top) MCP servers that provide:
  • Signal Discovery: Finding relevant signals (audiences, contextual, geographical, temporal) using natural language
  • Signal Activation: Pushing signals to decisioning platforms
  • Integration: Exposes data provider capabilities via MCP
Sales Agent (Right, Bottom) MCP servers that provide:
  • Media Product Discovery: Natural language inventory search
  • Media Execution: Creating and managing campaigns
  • Integration: Exposes publisher capabilities via MCP

Bottom Layer: Technical Execution

Agentic eXecution Engine (AXE) (Left) Real-time execution layer for:
  • Brand Suitability: Ensuring appropriate ad placement
  • Frequency Capping: Managing exposure limits
  • First-Party Data: Activating advertiser data
  • Dynamic Audience Targeting: Applying buyer segments at impression time
  • Integration: Connects via key-value pairs or RTB protocols
See AXE documentation for details. Decisioning Platform (Right) The technical infrastructure that:
  • Selects Impressions: Decides which ad to serve
  • Delivery Method: Direct campaigns or programmatic (RTB)
  • Examples: DSPs, SSPs, Ad Servers, Google Ad Manager, Kevel

Cross-Cutting: Governance & Human Oversight

Governance Agents Specialized agents that provide compliance and quality control across all layers:
  • Property Lists: Define where ads can and cannot run
  • Brand Suitability: Content filtering and suitability scoring
  • Quality Measurement: MFA score, ad density, consumer experience
  • Compliance: Privacy regulations (COPPA, TCF, GDPR)
  • Integration: Operates at setup time, real-time, and post-bid
See Governance Protocol for details. Human-in-the-Loop AdCP supports optional manual approval at key decision points:
  • Campaign Approval: Publishers can require human review before activation
  • Creative Review: Manual sign-off on ad content
  • Budget Thresholds: Automatic escalation above certain spend levels
  • Policy Exceptions: Human judgment for edge cases
This is not a real-time protocol - operations may take minutes to days when human approval is required.

The Complete AdCP Flow with Creative Agents

The AdCP ecosystem extends beyond media transactions to include creative production and delivery:

Three Interconnected Flows

Creative Flow (Top)
  • Creative Agent: AI-powered creative generation and optimization
  • Property: The publisher or platform displaying the ad
  • Consumer: The end user viewing the advertisement
Business Flow (Middle)
  • Advertiser: The brand with a message to deliver
  • Principal: The buying principal (agency, trading desk, or brand itself)
  • Media Seller: The supply-side entity selling ad inventory
Technical Flow (Bottom)
  • Brand Agent: Represents the advertiser’s objectives and brand guidelines
  • Media Agent: Orchestrates the media buying process
  • Sales Agent: Facilitates inventory discovery and transaction execution
  • Decisioning Platform: Makes real-time ad serving decisions
  • AXE (Agentic eXecution Engine): Provides real-time execution capabilities

Protocol Domains

AdCP covers multiple advertising domains:
DomainPurposeKey Tasks
Media BuyInventory discovery and purchasingget_products, create_media_buy
CreativeAd creative managementbuild_creative, sync_creatives
SignalsAudience and targeting signalsget_signals, activate_signal
Sponsored IntelligenceAI-powered conversational adssi_initiate_session, si_send_message

Quick Start

Client Libraries

JavaScript/TypeScript

npm install @adcp/client

Python

pip install adcp

Example: Natural Language Advertising

Instead of navigating multiple platforms, you can now say:
“Find audience signals of premium sports enthusiasts who would be interested in high-end running shoes, and activate them on Scope3.”
The AI assistant will:
  1. Search for relevant signals across connected platforms
  2. Show you options with transparent pricing
  3. Activate your chosen signals for use on decisioning platforms

For Platform Providers

AI is buying ads. Make sure it can buy yours. If you operate a data platform, DSP, or ad tech solution, AdCP lets AI agents discover and purchase your inventory. Review the Protocol Specifications to get started.

For Advertisers & Agencies

AI that sells products and builds brands. AdCP-enabled AI assistants can work across all your media partners through a single interface:
  1. Check if your platforms support AdCP
  2. Configure your AI assistant with AdCP-enabled platforms
  3. Start using natural language for your campaigns

Reference Implementations

Organizations

OrganizationPurpose
AgenticAdvertising.orgMembership community, working groups
AdCP (Ad Context Protocol)Protocol specification, schemas, documentation
AgenticAdvertising.org is the community. AdCP is the protocol. Members join AgenticAdvertising.org to develop and adopt AdCP.

Need Help?