Amazon Ads launches open beta of MCP Server to connect AI agents with ad workflows

Amazon Ads has opened the beta of its new Model Context Protocol (MCP) Server, introducing infrastructure designed to let AI agents interact directly with Amazon’s advertising systems through a single standardized connection.

The MCP Server is built on the open-source Model Context Protocol and acts as an intermediary layer between AI agents and Amazon Ads APIs. The goal is to reduce the need for custom, one-off integrations by translating natural language prompts into structured API calls that agents can reliably execute.

According to Amazon Ads, the server allows advertisers and ad tech partners to connect custom-built agents or third-party AI platforms such as ChatGPT, Claude, or Gemini to Amazon Ads features in minutes. Once connected, agents can manage campaigns, adjust budgets, retrieve performance reports, access billing data, and handle account-level settings without direct API orchestration by human operators.

The MCP rollout reflects broader industry efforts to make agentic AI practical inside advertising operations. While APIs remain central to ad tech stacks, Amazon Ads says they were not designed to support end-to-end autonomous workflows. Agents often struggle with context, outdated endpoints, or overly granular data, which can lead to inefficient or incorrect execution.

Amazon Ads positions the MCP Server as a “translation layer” that limits agents to current, supported APIs aligned with its internal domain model. This approach is intended to prevent common errors observed in internal testing, where agents either processed excessive data or relied on deprecated interfaces when performing tasks like attribution analysis or conversion reporting.

Alongside the open beta, Amazon Ads is introducing MCP “tools” that package multi-step advertising workflows into single executable actions. These tools are designed to function as instruction sets for agents, allowing them to complete tasks that typically require multiple API calls.

Examples include creating Sponsored Products campaigns end to end, expanding campaigns into new countries, shifting high-performing keywords across campaigns, or generating standard performance reports. These actions can be triggered through conversational prompts, with the MCP Server handling the underlying sequence of operations.

Amazon Ads says this structure reduces operational complexity while keeping agents focused on higher-level decision-making rather than basic system navigation.

The MCP Server grew out of internal challenges Amazon faced when testing AI agents for analytics and reporting. In several cases, agents produced technically correct results but did so inefficiently, such as querying years of historical data instead of using existing reports. These issues highlighted the need for clearer operational guidance rather than more sophisticated prompts.

By formalizing that guidance into tools and protocols, Amazon Ads aims to make agent-driven workflows more predictable and scalable for advertisers and partners.

The Amazon Ads MCP Server is now available globally in open beta to advertisers and partners with active Amazon Ads API credentials. The move places Amazon alongside other ad tech platforms working to standardize how AI agents interact with advertising systems, as the industry shifts away from manual execution toward autonomous and agent-assisted workflows.