Until now, fixing a broken programmatic campaign required manual checks across different platforms. With this new integration, the process is now automated and unified.
AdRoll and PubMatic have announced a partnership that uses the Model Context Protocol to connect their demand-side and supply-side platforms in a single diagnostic workflow. This lets AI agents find and fix campaign delivery issues in both systems in real time.
With this integration, AdRoll’s agents can directly check PubMatic’s deal diagnostics. They can quickly find the root causes of delivery problems, such as pacing limits, creative issues, or publisher-side factors, and suggest solutions without manual checks across different platforms. The companies say this can reduce issue resolution from days to just minutes.
“Advertising doesn’t operate in silos, and the systems behind it can’t either,” said Vibhor Kapoor, chief executive of AdRoll. “Our strategy is to build a more connected and flexible ecosystem where AI can work across platforms, with the ability to incorporate new data sources and technologies as they emerge.”
The integration builds on AdRoll’s investment in MCP as an open standard for connecting AI systems to marketing workflows and validates PubMatic’s first agent-to-agent MCP specification, published in September 2025. The companies say the approach demonstrates that independent platforms can interoperate without tightly coupled integrations — a flexibility they argue is essential as deal-based buying strategies grow more complex.
Also Read: Push Notifications Are Broken. Here Is What Comes After Them
“One of the biggest challenges in traditional programmatic workflows is a lack of visibility into how demand-side decisions impact campaign delivery,” said Alex Shephard, VP of advertiser solutions at PubMatic. “This collaboration shows that intelligence across the supply chain drives better outcomes, giving marketers and partners a clearer view into what’s happening and why.”
The partnership reflects a broader industry push toward interoperable, agent-driven systems. As open standards like MCP gain wider adoption, the companies argue that direct agent-level collaboration between DSPs and SSPs will become operationally standard rather than experimental.