AdRoll and PubMatic on April 23, 2026, announced an integration that connects AI agents across their respective demand-side and supply-side platforms, using the Model Context Protocol (MCP) as the connective layer. The integration targets one of programmatic advertising's most persistent operational frustrations: when a deal stops delivering, figuring out why has historically required manual log exports, support tickets, and days of back-and-forth between DSP and SSP teams. This partnership attempts to compress that process to minutes.

The announcement, made by AdRoll from San Francisco and distributed via GlobeNewswire, describes a working implementation rather than a roadmap. According to AdRoll and PubMatic, the integration is already functional: AdRoll's AI agents can query PubMatic's deal diagnostics in real time, pull root-cause analysis on delivery failures, and surface recommended corrective actions - all within a single coordinated workflow that does not require logging into a second platform or waiting for a human counterpart to respond.

What the integration actually does

The technical premise is straightforward, even if its implications for how programmatic operations teams work are significant. When a deal underdelivers, the causes can sit on either side of the transaction: a pacing constraint on the demand side, a creative block, a publisher-side floor price mismatch, or a targeting configuration that is too narrow to clear. Identifying the correct cause has previously meant correlating data across two separate systems, each with its own reporting structure and vocabulary.

According to the press release, AdRoll's agents now query PubMatic's agentic diagnostics directly through MCP. The protocol functions as a standardized communication layer between the two platforms, allowing AdRoll's system to issue a diagnostic request and receive a structured response from PubMatic's infrastructure - without custom API work or tightly coupled integration. The DSP-side agent issues the query; the SSP-side agent responds with diagnostic information. Both continue operating within their respective platforms.

Vibhor Kapoor, chief executive officer at AdRoll, was quoted in the announcement: "Advertising doesn't operate in silos, and the systems behind it can't either. 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."

Alex Shephard, vice president of advertiser solutions at PubMatic, described what the integration addresses from the supply side: "In traditional programmatic workflows, one of the biggest challenges is a lack of visibility into how demand-side decisions impact campaign delivery. 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."

MCP's role as the connective layer

The Model Context Protocol was originally developed by Anthropic and subsequently donated to the Linux Foundation. It operates as a standardized communication framework between AI applications and external systems - comparable in concept to a universal connector that allows large language models to access APIs, databases, and tools through a consistent interface rather than a bespoke connection for each.

In the programmatic advertising context, MCP's practical advantage is what it avoids. A conventional DSP-SSP integration requires custom API development, agreed-upon data schemas, and continuous engineering maintenance from both parties whenever one platform changes its data structure. According to the AdRoll-PubMatic announcement, their integration achieves interoperability without any of that overhead.

PubMatic published its first agent-to-agent MCP specification in September 2025. The AdRoll integration validates that specification against a live operational use case. The Ad Context Protocol launched on October 15, 2025, with PubMatic as one of six founding companies alongside Scope3, Yahoo, Swivel, Triton Digital, and Optable - an earlier attempt to standardize how AI agents discover inventory, compare pricing, and activate campaigns across platforms. The AdRoll integration operates on MCP directly and does not require AdCP, though both share the same foundational protocol layer.

The adoption of MCP across advertising infrastructure has been rapid. Google released an open-source MCP server for its Ads API on October 7, 2025, enabling AI applications to query advertising campaign data through natural language. Microsoft launched its Clarity MCP server in June 2025. Amazon Ads announced a closed beta for its own MCP Server on November 13, 2025. The IAB Tech Lab's Agent Registry held 10 active MCP-registered entries as of March 11, 2026, all operating under the Model Context Protocol standard, with PubMatic holding registry entry GPP and TCF ID number 76.

The IAB Europe demonstration

The integration did not emerge from a laboratory. In April 2026, a live demonstration at an IAB Europe webinar showed PubMatic's MCP server being used to troubleshoot a deal in real time. According to details documented by PPC Land, the system analyzed 1,200 incoming bids and 2,000 PubMatic winning bids, and identified a 1% DSP bid adherence rate as the root cause of poor delivery - all within seconds. That demonstration preceded the AdRoll announcement by roughly two weeks and established that the underlying diagnostic infrastructure functions under live conditions.

The AdRoll integration extends that capability beyond an internal PubMatic demonstration into a cross-company, cross-platform workflow. Two independent organizations, with no shared ownership structure, are exposing their internal systems to each other's AI agents through an open protocol.

PubMatic's AgenticOS context

The deal troubleshooting capability that AdRoll now accesses is part of PubMatic's broader AgenticOS, launched on January 5, 2026. PubMatic describes AgenticOS as an operating system layer for autonomous advertising execution across premium digital environments. Early testing on that infrastructure demonstrated campaign setup time reductions of 87% and issue resolution improvements of 70%, according to PubMatic's own performance metrics at the time of the launch announcement.

AgenticOS has been the platform through which PubMatic has pursued several integrations in 2026. In March 2026, Optable's Audience Agent was embedded into PubMatic AgenticOS, enabling buyers to discover and activate publisher first-party audience data without that data leaving the publisher's environment. The AdRoll integration represents a different type of expansion: from supply-side audience capabilities into cross-platform diagnostic workflows initiated from the demand side.

The December 2025 agentic campaign executed by PubMatic with independent agency Butler/Till and Geloso Beverage Group's Clubtails brand, using a Claude-based agent communicating with AgenticOS infrastructure, produced 40% more impressions than comparable non-agentic activity and a cost efficiency of 5.5x for Butler/Till, with a 98% average video completion rate - figures cited by PubMatic in context of that test's performance.

Why this matters for programmatic operations

The operational significance of the AdRoll-PubMatic integration sits in a specific and common workflow: deal-based buying. As advertisers have moved toward more structured private marketplace and programmatic guaranteed arrangements - rather than open auction buying - the complexity of diagnosing delivery failures has increased. Deals involve bilateral configurations: the buyer sets targeting and bid parameters, the seller sets floors and eligibility rules, and mismatches between the two can produce delivery failures that neither side can diagnose from its own data alone.

PPC Land has tracked the expansion of MCP-based advertising tools throughout 2025 and into 2026, across Yahoo, Amazon, FreeWheel, PubMatic, AdRoll, Dstillery, and others. The pattern across these implementations is consistent: platforms are embedding AI into transactional and diagnostic workflows, not only into reporting interfaces. The AdRoll-PubMatic integration is notable because it crosses company boundaries - most agentic implementations to date have operated within a single platform or platform family.

Yahoo DSP launched agentic AI capabilities on January 6, 2026, including a troubleshooting agent that proactively identifies pacing and delivery issues within the Yahoo DSP ecosystem. That capability is internal to Yahoo. The AdRoll-PubMatic implementation extends a comparable diagnostic function across an external DSP-SSP relationship - a distinction that carries engineering and commercial implications.

What the integration does not yet do

The scope of the announced integration is deliberately specific. According to the press release, AdRoll's agents can query PubMatic's deal diagnostics and receive recommended actions. The announcement does not describe automated remediation - the execution of fixes without human review. Surfacing recommended actions is different from taking them autonomously. That distinction matters for how advertising teams will actually use the workflow: a human campaign manager receives a diagnosis and recommendation within a single interface, rather than conducting that investigation manually, but the decision to act remains with the person.

The announcement also does not specify production scale, the number of campaigns currently operating within this diagnostic framework, or the full range of deal types covered. The press release describes the integration as a demonstration of how cross-platform agentic communication can work, rather than a mature product with published adoption figures.

Industry skepticism about MCP-based advertising protocols has been present since the Ad Context Protocol launch. An industry veteran argued on October 15, 2025, that AdCP was "the tail wagging the dog", pushing automation tools before fixing the structural problems in programmatic supply chains. The AdRoll-PubMatic integration addresses a narrower problem - deal troubleshooting rather than full campaign automation - which may insulate it from those broader criticisms, at least in its current form.

AdRoll's broader MCP positioning

The collaboration builds on AdRoll's early investment in MCP as an open standard for connecting AI systems to marketing workflows. According to the press release, the integration validates PubMatic's first agent-to-agent MCP spec published in September 2025 - meaning AdRoll is not only adopting MCP as an end user but is participating in validating how the protocol behaves in cross-platform conditions.

AdRoll has pursued several integrations in the past year. The platform reached 200 brands activating Experian audiencesacross web and connected TV campaigns as of October 2025, following the partnership announced in April 2025. In October 2025, AdRoll integrated HUMAN Security for fraud defense and viewability tracking, becoming the first demand-side platform to adopt HUMAN's viewability solution alongside real-time fraud defense, with invalid traffic detection operating within 12 milliseconds. The MCP integration with PubMatic is the first publicly announced cross-platform agent-to-agent capability for AdRoll.

AdRoll, a NextRoll brand, has operated as a multi-channel advertising platform primarily serving mid-sized businesses. The platform processes identity data, behavioral insights, and performance metrics across display, native, video, mobile, social, and connected TV environments. AdRoll's State of Digital Advertising Report, published on March 24, 2026, documented display retargeting CPMs surging 18% year-over-year in January and February 2026, while prospecting CPMs fell 11% - figures based on data from more than 20,000 online businesses.

PubMatic (NASDAQ: PUBM) is listed on the Nasdaq and has operated as a supply-side platform since 2006. The company describes itself as the leading AI-powered ad tech company delivering digital advertising performance. PubMatic registered its MCP server with the IAB Tech Lab Agent Registry, carrying GPP and TCF ID number 76 - a detail that links its protocol registry entry to its Global Privacy Protocol registration and connects privacy compliance infrastructure to the agentic discovery layer.

The initiative reflects a broader industry shift toward more interoperable, agent-driven systems. As open standards like MCP mature, they enable DSPs and SSPs to collaborate directly at the agent level - moving beyond siloed platforms and manual handoffs. This integration demonstrates that DSP-SSP collaboration is operationally viable in an agentic context. As more platforms adopt these standards, the ecosystem may benefit from faster issue resolution and reduced operational friction across the supply chain.

Timeline

Summary

Who: AdRoll, an AI-powered multi-channel advertising platform based in San Francisco and a NextRoll brand, and PubMatic (NASDAQ: PUBM), a Nasdaq-listed supply-side platform operating since 2006. The announcement cited Vibhor Kapoor, CEO of AdRoll, and Alex Shephard, VP of Advertiser Solutions at PubMatic.

What: An agent-to-agent integration using the Model Context Protocol, allowing AdRoll's demand-side AI agents to query PubMatic's supply-side deal diagnostics in real time. The integration identifies root causes of delivery failures - including pacing constraints, creative blocks, and publisher-side factors - and surfaces recommended actions within a single coordinated workflow, reducing resolution time from days to minutes.

When: The integration was announced on April 23, 2026. It builds on PubMatic's agent-to-agent MCP specification first published in September 2025, and on PubMatic's AgenticOS infrastructure launched January 5, 2026.

Where: The integration connects AdRoll's demand-side platform with PubMatic's supply-side platform through the Model Context Protocol, a standard originally developed by Anthropic and donated to the Linux Foundation. No shared infrastructure or proprietary coupling is required between the two platforms.

Why: Deal-based programmatic buying creates delivery failures that span DSP and SSP boundaries, making root-cause diagnosis difficult without access to both platforms' data simultaneously. The integration addresses that visibility gap by enabling an AI agent on the demand side to query supply-side diagnostics directly - without manual investigation across multiple platforms, support tickets, or custom API development.

Share this article
The link has been copied!