A media buyer managing a streaming campaign today spends hours pulling performance reports, checking whether ad pods are running too long, flagging frequency problems, and renegotiating placements that aren't delivering. Most of that work is reactive. By the time the data arrives, the impressions have already run. FreeWheel wants to eliminate that loop entirely.
How It Works
Model Context Protocol (MCP) is an open standard that lets AI agents talk to external systems in a structured way. Think of it as a universal adapter. Instead of building custom integrations for every platform, an AI agent that speaks MCP can connect to any system running an MCP server and immediately understand what data is available and what actions it can take.
FreeWheel built an MCP server on top of its ad platform. That server exposes campaign data, deal performance, inventory quality signals, spend tracking, and commitment monitoring to any AI agent that connects to it. Before this, getting that information required a human logging into a dashboard, pulling a report, reading it, and deciding what to do. Now a machine can ask the same questions and get structured answers.
PMG, an independent media agency, is the first to pilot the system. Their operating system, Alli, connects directly to FreeWheel's MCP server. Alli can query deal performance, check inventory quality, monitor pacing, and act on the answers without waiting for a human to review a spreadsheet and send an email.
Why the Foundation Matters
FreeWheel's Viewer Experience Lab has spent years quantifying what makes ad breaks work. Pods under two minutes double brand recall. Consistent ad lengths within a pod improve sentiment by 7%. Frequency caps at two to three exposures lift purchase intent. Publishers had the playbook. Most didn't execute it.
Now the execution is automated. The intelligence tools sitting on top of the MCP server can apply those rules to live campaigns. The research wrote the playbook. The infrastructure enforces it.
Most AI in ad tech is scaling against unverified data. FreeWheel's system optimizes against something it controls: its own research on what makes ad breaks perform.
🆘 SOS Insight: FreeWheel researches which ad designs work, builds the infrastructure that automates them, and operates under Comcast, whose parent company owns significant premium inventory.
The Bigger Picture
FreeWheel's MCP server is one node. The Interactive Advertising Bureau (IAB) Tech Lab's Agent Registry, launched weeks earlier, is building the directory. The registry is itself an MCP server, meaning an AI agent can query it to discover other agents across the ecosystem: demand-side platforms, identity resolution vendors, measurement providers. If FreeWheel built the first room, the IAB is building the hallway that connects all the rooms together.
This matters because the value of any single MCP server multiplies when agents can find each other, verify capabilities, and transact without human intervention. The agentic transaction layer for premium video is being wired together now.
What It Means
Ad buyers should ask whether their current partners offer MCP-level integration or still rely on dashboard-and-email workflows. The gap between automated decisioning and manual optimization gets wider every quarter.
Publishers who ignored the Viewer Experience Lab findings now face a system that implements them by default. The research is optional reading. The infrastructure makes it mandatory.
Ad tech vendors without an agentic strategy face a discovery problem. If your platform isn't registered, queryable, and interoperable, an AI agent can't find you. Increasingly, the AI agent is the buyer.
This Article By the Numbers
- 700 viewers participated in FreeWheel's Viewer Experience Lab study with MediaScience
- Pods under 2 minutes doubled brand recall compared to 3-minute pods
- Consistent ad lengths within a pod improved viewer sentiment by 7%
- Frequency caps at 2 to 3 exposures per program lifted both ad recognition and purchase intent
.png)