Amazon launches closed beta for AI agent advertising integration

Amazon Ads announced on November 13, 2025, the launch of a closed beta for its Model Context Protocol Server, a standardized access layer designed to connect artificial intelligence models and agents with Amazon's advertising platform. The MCP Server transforms complex API operations into conversational queries, enabling large language models to access campaign data, performance metrics, billing information, and account details through natural language interactions.

The Amazon Ads MCP Server operates as an intermediary layer between AI applications and Amazon's advertising infrastructure. According to the technical documentation, the server "transforms complex multi-field API operations into simple conversational queries, making Amazon Ads insights like campaigns, performance metrics, billing, and account information accessible to Large Language Models (LLMs) and AI applications through natural language interactions."

Model Context Protocol represents an open-source standard developed by Anthropic that enables developers to build secure, two-way connections between data sources and AI-powered applications. The protocol establishes a framework for how AI applications and external systems interact, specifying types of contextual information that can be shared and the range of actions that can be performed.

The architecture includes three primary participants: the Amazon Ads MCP Server, which standardizes and exposes Amazon Ads features through tools, resources, and prompts; the MCP Client, which manages secure data exchange between AI applications and the server; and the primitives that define fundamental building blocks for interactions. Amazon Ads currently supports tools as primitives, with resources and prompts planned for future releases.

Tools represent functions exposed to agents with descriptions, input properties, and return values that allow agents to perform actions. Resources provide contextual information to AI applications such as file contents, database records, and API responses. Prompts serve as reusable templates that structure interactions with language models.

The documentation identifies several example use cases. Users can query campaign performance with requests like "Show me campaign performance for October 2025 on [account_id]," access account information by asking "Show me a list of all of my advertising accounts," generate reports with requests such as "Create a Campaign report for [account_id]," and update campaign settings through simple instructions like "Increase my campaign budget to $500 on [campaign_id]."

Implementation requires specific technical prerequisites. According to the setup documentation, each request to the Amazon Ads MCP server requires the Amazon-Ads-ClientId header containing the client identifier of a Login with Amazon application authorized to access the MCP Server, and an Authorization header containing the string "Bearer" prepended to an access token representing the application's permission to access data and services for a given Amazon user.

The system operates across three regional endpoints. North America uses https://advertising-ai.amazon.com/mcp, Europe utilizes https://advertising-ai-eu.amazon.com/mcp, and the Far East region connects through https://advertising-ai-fe.amazon.com/mcp. These URLs are not region-agnostic, requiring advertisers to specify the correct endpoint for their geographic location.

Account identification differs from traditional Amazon Ads API implementation. While the Amazon Ads API requires accountIds in headers or fields such as inserting a profileId into the Amazon-Advertising-API-Scope header, the MCP server handles this differently. The server requires account identifiers as parameters in the request body rather than headers, aligning with MCP protocol standards while maintaining equivalent authorization and account scoping functionality.

The documentation lists three identifier types: profileId for sponsored ads accounts, managerAccountId for manager accounts, and advertiserAccountId which can represent a DSP advertiser account, global account ID, or Amazon Marketing Cloud instance ID depending on the specific tool being called.

Advertisers can connect to the MCP server using Claude from Anthropic, ChatGPT from OpenAI, Amazon Q, Amazon Bedrock, Amazon AgentCore, and other MCP-compatible applications. According to the documentation, "Amazon Ads does not own or operate the LLMs or AI applications that connect through the MCP Server." The company maintains and updates the server so developers and AI applications can consistently access the latest functionality without re-engineering integrations.

The system implements tool filtering capabilities allowing clients to restrict agent interactions. The server supports filtering by read-only status through the readOnlyHint annotation, enabling administrators to prevent write operations when necessary. Additionally, the architecture supports filtering by tool groups, which organize functions into four main categories: account_management for account creation and configuration, billing for invoices and payment data, campaign_management for advertising campaigns across product types, and reporting for performance analytics.

Tool naming follows a standardized convention using the format <tool_group>-<tool_name>, where the tool group represents the functional category and the tool name describes the specific operation. For example, account_management-create_advertiser_account indicates a tool in the account management group that creates advertiser accounts. This naming structure enables efficient tool discovery using regex patterns based on tool groups.

The announcement arrives as Amazon expands its API infrastructure throughout 2025. Amazon launched a unified reporting system on November 11, 2025, consolidating reporting interfaces across the platform. The company introduced a unified account structure in October 2025 enabling global campaign management across advertising products.

Amazon also released an AI agent for automated campaign management on November 12, 2025, which operates within existing Amazon Ads interfaces. The agent provides three capabilities: creating campaigns from brief descriptions, providing intelligent targeting recommendations by analyzing audience segments, and accelerating analytics workflows in Amazon Marketing Cloud by translating business questions into SQL queries.

The MCP Server represents Amazon's entry into standardized AI agent protocols for advertising. Six advertising technology companies launched the Ad Context Protocol on October 15, 2025, built on Anthropic's Model Context Protocol. That framework enables AI agents to discover inventory, compare pricing, and activate campaigns across different advertising platforms. Major platforms including Google, Trade Desk, and Amazon DSP have not yet adopted Ad Context Protocol.

Google released an open-source MCP server on October 7, 2025, enabling large language models to connect with Google Ads API for read-only reporting and diagnostics through natural language queries. The company unveiled comprehensive AI advertising tools at Think Week 2025 on September 10, introducing three AI-powered advisory systems described as "agentic capabilities."

For marketing professionals managing Amazon advertising campaigns, the MCP Server introduces programmatic access to campaign operations through conversational interfaces. Tasks that previously required navigating platform interfaces, managing spreadsheets, and technical API expertise become accessible through natural language queries processed by AI agents.

The closed beta status indicates limited availability. According to a LinkedIn post by Alexander Brockhoff, Product Manager for Amazon Ads Developer Experience, interested parties can request access through Amazon's developer portal. The announcement states the server is "Enabling partners to build smarter ads agents" and encourages feedback from beta participants.

Implementation considerations include authentication workflows, error handling for API responses, and integration with existing campaign management systems. The asynchronous nature of some operations requires developers to implement proper response handling for long-running requests such as report generation.

The technical documentation provides comprehensive error codes. Bad requests return 400 status codes, authentication failures generate 401 responses, rate limiting manifests through 429 status codes, content size restrictions enforce 413 responses, and server-side errors display as 500, 502, 503, or 504 codes depending on specific infrastructure issues.

Amazon's approach differs from traditional API access by abstracting technical complexity behind natural language interfaces. Instead of constructing JSON payloads with specific field requirements and navigating multiple endpoints, users describe desired actions conversationally. The MCP Server interprets these descriptions, translates them into appropriate API calls, executes operations, and returns results in formats accessible to AI applications.

The system supports Amazon DSP campaigns, sponsored ads products, Amazon Marketing Cloud analytics, and account management functions. This comprehensive coverage enables agents to perform most advertising operations without requiring separate API integrations for different product lines.

Connection security follows Amazon's standard authentication protocols using Login with Amazon credentials. All communications between MCP clients and the server occur over HTTPS, with access tokens providing user-specific authorization scopes. The architecture ensures that agents can only access data and perform operations permitted by the authorizing user's account privileges.

Amazon maintains the MCP Server infrastructure, handling updates as the underlying Amazon Ads API evolves. This approach reduces maintenance burden for developers building AI-powered advertising tools, as changes to API specifications propagate through the MCP Server without requiring client-side modifications.

The beta program provides Amazon with user feedback on tool design, response formats, and capability priorities before general availability. Beta participants shape the final implementation through practical experience identifying gaps, workflow inefficiencies, and feature requests that inform Amazon's development roadmap.

As advertising technology increasingly incorporates AI agents for campaign management, standardized protocols enable interoperability across platforms and AI applications. Amazon's adoption of Model Context Protocol signals alignment with industry standards rather than proprietary approaches, potentially accelerating ecosystem development around AI-powered advertising automation.

Timeline

• September 10, 2025: Google unveils agentic capabilities at Think Week 2025, introducing three AI-powered advisory systems for campaign management
• October 7, 2025: Google releases open-source MCP server for Ads API integration enabling natural language queries for reporting
• October 15, 2025: Six advertising technology companies launch Ad Context Protocol built on Anthropic's Model Context Protocol
• October 2025: Amazon DSP forecasting API reaches general availability for programmatic campaign projections
• October 2025: Amazon launches unified account for global campaign management across advertising products
• November 11, 2025: Amazon announces unified reporting experience in open beta consolidating reporting interfaces
• November 12, 2025: Amazon launches AI agent for automated campaign management with natural language interfaces
• November 13, 2025: Amazon Ads announces MCP Server closed beta enabling AI agents to access advertising platform through standardized protocol

Summary

Who: Amazon Ads announced the MCP Server for advertisers worldwide with Amazon DSP accounts, sponsored ads accounts, and Amazon Marketing Cloud access. Developers can connect through Claude (Anthropic), ChatGPT (OpenAI), Amazon Q, Amazon Bedrock, Amazon AgentCore, and other MCP-compatible applications.

What: The Amazon Ads MCP Server is a standardized access layer that transforms complex multi-field API operations into conversational queries. The server exposes Amazon Ads features through tools, resources, and prompts, enabling AI agents to query campaign performance, access account information, generate reports, and update campaign settings through natural language interactions. The system currently supports tools as primitives, with six endpoints organized into four functional groups: account management, billing, campaign management, and reporting.

When: Amazon announced the MCP Server closed beta on November 13, 2025. The announcement follows Amazon's November 12, 2025 launch of an AI agent for automated campaign management and November 11, 2025 introduction of a unified reporting experience.

Where: The MCP Server operates across three regional endpoints: North America (https://advertising-ai.amazon.com/mcp), Europe (https://advertising-ai-eu.amazon.com/mcp), and Far East (https://advertising-ai-fe.amazon.com/mcp). The server provides access to Amazon's global advertising infrastructure including Amazon DSP, sponsored ads products, and Amazon Marketing Cloud analytics.

Why: The MCP Server addresses technical complexity in Amazon's advertising APIs by enabling natural language interactions rather than requiring detailed knowledge of endpoint structures, parameter formats, and response parsing. The standardized protocol enables developers to build AI-powered advertising tools that stay current with platform changes without re-engineering integrations, as Amazon maintains the server infrastructure and updates functionality automatically.