Home Marketers Amazon Ads Opens A Beta Test For Its New MCP Server

Amazon Ads Opens A Beta Test For Its New MCP Server

SHARE:
"Model Context Protocol" handwritten on a post-it note, stuck on a bulletin board

APIs are so 2024.

The new hot trend is what’s known as Model Context Protocol, or MCP.

And Amazon Ads is entering the fray with a built-in MCP product meant to be a first foot in the door for advertisers using agentic systems.

On Monday, Amazon Ads opened a beta program for its new MCP server that connects preexisting software systems and their APIs to agentic AI systems.

Amazon Ads hosts the infrastructure on behalf of advertisers, said Paula Despins, Amazon Ads’ VP of ads measurement. The Amazon Ads MCP will also be natively integrated with Amazon campaign management tools, she added, though on the AI side clients can bring in their own LLM or agentic AI solution.

Lost in translation

The new Amazon Ads MCP acts as a “translation layer” between the software and the agents without having to introduce custom integrations for each one, said Despins.

The idea isn’t a fundamental overhaul of how an online ad campaign works so much as a way to reduce human work hours and general frustration because their new software services don’t sync nicely. For example, through the beta program, advertisers will now automatically have access to new features that are rolled out through the MCP server.

Working with an MCP removes a lot of “heavy lifting” that’s part of advertisers’ current processes, Despins added, since there isn’t a new integration for each individual API.

The new server includes tools that respond to natural language prompts, simplifying processes that would otherwise take several API calls, and perhaps even a data analyst or someone with knowledge of a programming language like SQL.

The tools function sort of like dominoes: A user prompts a tool, which then provides the MCP server with specific instructions. From there, the server executes the instructions by connecting to Amazon Ads’ APIs.

The new additions were designed for tasks that advertisers perform most often.

For instance, one tool can develop a full campaign with ad groups, audiences and a budget breakdown in response to a prompt that just lists the product and budget, Despins said. Another tool shifts high-performing keywords from one campaign to many others.

Without MCPs to translate “very granular APIs,” agents often make mistakes due to an overabundance of information, she said. “It’s the equivalent of just dumping the file cabinet out instead of saying, ‘Here’s the three files you need.’”

(Instruction) manual labor

The new MCP beta program was developed in response to problems that Amazon faced itself, not just with its Amazon Ads clientele, said Despins.

Amazon Ads found that when its execs asked for insights on a conversion report, the existing agent might go to Amazon Marketing Cloud and analyze three years of data, rather than looking at the standard report it was given and expected to use for the analytics.

“It actually did a pretty good job,” Despins said, “but it was a lot slower.”

The initial solution was simply to prompt the agent more effectively and specifically for online ad campaign functions. The tactic … sort of worked, according to Despins. Except that the agent used a deprecated version of an API from three years prior.

The new MCP server is built only on the APIs that are updated to match Amazon Ads’ “domain model standard,” said Despins, so agents will only refer to the most up-to-date versions of a given API.

“A human would probably look a little harder before it made those missteps,” said Despins.

Which, she said, then prompted the question: “Why not give the agent the instruction manual?”

Must Read

Scales and hands touching the bowls with index fingers from opposite sides. Arguments, evidence and tricks in trial. Concept of judging, trial and justice

The FTC Bars Kochava From Selling Sensitive Data Without Consent

It’s been nearly four years since the Federal Trade Commission first accused Kochava of selling highly sensitive location data. Now, the two have finally reached a settlement.

Comic: CTV Tracking

Upfronts Advertisers Say They Want Outcomes – And Amazon Licks Its Chops

Amazon has packaged a handful of upgrades to its ads measurement solutions, obviously catered to TV and streaming media advertisers.

AdExchanger Senior Editors Anthony Vargas and Alyssa Boyle.

POSSIBLE 2026: AdExchanger's Hot Takes

AdExchanger Senior Editors Alyssa Boyle and Anthony Vargas share their takeaways from three days chatting about agentic AI at POSSIBLE.

Privacy! Commerce! Connected TV! Read all about it. Subscribe to AdExchanger Newsletters

Reddit Reports A 75% Boost In Q1 Ad Revenue As It Reaches For 100 Million Daily US Users

Generative AI search has pushed traffic off a cliff across most of the internet, but not on social platforms. Reddit included.

POSSIBLE 2026: Can AI Help Agencies Finally Break Down Those Silos?

Domenic Venuto, indie agency Horizon Media’s chief product and data officer, sat down with AdExchanger during POSSIBLE at the Fontainebleau in Miami to unpack the role of AI in today’s media and advertising landscape.

Google Touts Its AI Ad Tech Adoption And New AI Max Features

Google announced new features and ad types for AI Max, its AI-based bidding product for search and shopping or sponsored product ads. The company also touted “hundreds of thousands” of advertisers using AI Max.