Amazon opens access to its AI agent advertising platform via MCP

Amazon has launched an open beta of its Amazon Ads Model Context Protocol (MCP) server, enabling advertisers and partners to connect AI agents to Amazon Ads through a single standardized integration instead of juggling multiple custom connections. The company unveiled the solution at the annual IAB meeting with the goal of simplifying tasks like campaign setup, budget management, and report generation by acting as a translator between natural language and the Amazon Ads API.

The technology is built on Model Context Protocol, originally developed by Anthropic, and features “tools for common actions” that consolidate multi-step advertising workflows into a single command set for the agent. Amazon is betting on convenience and speed – while tightening its control over how partners integrate with its ad services.

how amazon MCP works

MCP acts as a middleware layer: the AI agent sends natural language requests, MCP translates these into structured Amazon Ads API calls, and activates predefined “tools” designed for typical scenarios. Paul Despins, VP of ad measurement at Amazon Ads, explains this reduces the “reasoning load” on agents-they no longer have to figure out which API version to use or the specific steps each endpoint requires.

• One integration replaces many custom connectors;
• Natural language queries are converted into structured API calls;
• Toolkits handle common multi-step tasks like campaign creation, budget adjustments, and reporting;
• Built-in safeguards prevent typical agent errors, such as using outdated API versions or unnecessary data reprocessing.

Amazon shared examples from internal tests: one agent tasked with building a conversion path wrote its own code and processed three years of data via Amazon Marketing Cloud-something that could have been avoided with a clear set of MCP instructions. These scenarios motivated Amazon to create “instructions” for widely used workflows.

winners and losers

The winners are practical players: advertisers without large engineering teams, agencies adopting agent workflows quickly, and platforms that can sell more automated campaigns. Small and medium advertisers finally gain access to automation that previously demanded costly integrations.

The losers? Those thriving on custom integrations: independent integrators and third-party connector developers may lose business. Additionally, partner independence takes a hit-the more actions routed through Amazon’s standardized layer, the more partners become dependent on its rules and API versions.

what it means for the ad market

MCP isn’t Amazon’s only attempt to standardize agent interactions in advertising; similar efforts like AdCP are also emerging in the ecosystem. For the market, this signals a rapid shift from routine manual tasks to agent-driven automation-and a redistribution of value towards the platforms that set standards.

Amazon has strong credentials: its ad business grew rapidly last year – advertising revenue rose 24% in Q3 to $17.7 billion. If MCP really does simplify campaign scaling, it’ll strengthen Amazon’s commercial edge over rival platforms.

But risks lurk-security, auditability of agent decisions, and transparency. Agents granted control over budgets and targeting must remain under human oversight. The question of “who’s responsible if an agent makes a wrong call” will become as much policy as technology.

what’s coming next

Expect quick adoption shifts: platforms will either support MCP compatibility or push their own standards. Ad agencies will massively test agent-driven scenarios, while regulators and big clients demand reporting and rollback mechanisms. The question isn’t if automation will rise – it already has. It’s who will write the rulebook and whether the ecosystem can maintain a diverse set of solutions.