AdRoll and PubMatic on April 23, 2026, announced an integration that allows AI agents on the demand side to query supply-side diagnostics directly - using the Model Context Protocol as the connective layer - in what the companies describe as one of the first demonstrated implementations of cross-platform agent-to-agent communication in programmatic advertising.

The announcement is specific in its scope. This is not a roadmap or a framework proposal. According to the companies, AdRoll's agents can already query PubMatic's deal diagnostics to identify root causes of delivery issues, including pacing constraints, creative blocks, and publisher-side factors, and surface recommended actions. The practical implication: teams can move from detecting a problem to resolving it in minutes rather than days, without manually exporting logs or reconciling reports across multiple platforms.

That claim matters in an industry still working out what "agentic advertising" actually means in production. PPC Land has tracked the rapid expansion of MCP-based advertising tools throughout 2025 and into 2026, with most implementations remaining either within a single platform or at the level of natural-language query interfaces for reporting data. Cross-platform agent communication - where a DSP-side agent sends a structured request to an SSP-side agent and receives actionable diagnostic output - is considerably harder, and considerably rarer.

What MCP actually does here

The Model Context Protocol was originally developed by Anthropic and donated to the Linux Foundation. It functions as a standardized communication framework between AI applications and external systems - comparable to a universal connector that allows large language models to access APIs, databases, and tools through a consistent interface. The protocol operates through a client-server model where host applications maintain connections to multiple servers, each exposing specific capabilities or data access.

In the advertising context, MCP's relevance lies in what it avoids: tightly coupled integrations. Traditionally, when a demand-side platform needs information from a supply-side platform, it involves custom API work, agreed-upon data formats, and ongoing engineering maintenance on both sides. According to the announcement, the AdRoll-PubMatic integration achieves interoperability without any of that coupling. The DSP-side agent issues a query; the SSP-side agent responds with diagnostic information; both operate within their respective platforms.

PubMatic published its first agent-to-agent MCP spec in September 2025. The AdRoll integration validates that spec against a real workflow - a demand-side operator troubleshooting a live deal using supply-side data, without switching platforms or waiting for a support ticket.

According to Vibhor Kapoor, chief executive officer at AdRoll: "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. This integration is a terrific example of how that comes to life, helping marketers understand and act on what's happening in their campaigns."

Alex Shephard, vice president of advertiser solutions at PubMatic, addressed the structural problem this solves: "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. It's an important step toward more transparent, coordinated workflows across the advertising ecosystem."

The diagnostic workflow

The integration targets a specific operational pain point: deal troubleshooting. In programmatic advertising, deals - particularly private marketplace and programmatic guaranteed transactions - require configuration on both the DSP and SSP side. When delivery fails or underperforms, diagnosis typically requires investigating settings on both sides simultaneously. That has historically meant manual back-and-forth: a trader opens a ticket with the SSP, waits for a response, receives log data in a format that doesn't align with their DSP's reporting structure, and tries to reconcile the two.

The MCP workflow replaces that process with a single coordinated diagnostic flow. The DSP-side agent evaluates campaign configuration - targeting parameters, bid logic, pacing settings, creative specifications - while simultaneously querying PubMatic's agentic diagnostics for exchange-level information. Both data sets are evaluated together in real time. Root causes that emerge from this joint view include pacing constraints set at the deal level, creative formats blocked by publisher-side rules, or supply-side factors affecting win rates.

PubMatic launched its AgenticOS in January 2026, describing it as an operating system built 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 performance metrics at the time. The deal troubleshooting capability now demonstrated with AdRoll connects to that same diagnostic infrastructure, exposing it to an external DSP partner through a standardized protocol rather than a proprietary integration.

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: the system analyzed 1,200 incoming bids, 2,000 PubMatic winning bids, and identified a 1% DSP bid adherence rate as the root cause of poor delivery - all within seconds. The AdRoll integration extends this type of diagnostic capability into a two-way, cross-platform workflow rather than an internal one.

Context: where this fits in the MCP timeline

The industry's adoption of the Model Context Protocol in advertising has moved quickly. Microsoft launched a Clarity MCP server in June 2025 for web analytics queries. Google released an open-source MCP server for its Ads API on October 7, 2025, enabling natural language queries against advertising campaign data. Amazon Ads announced a closed beta for its own MCP Server on November 13, 2025, transforming complex API operations into conversational queries.

On October 15, 2025, six companies - Scope3, Yahoo, PubMatic, Swivel, Triton Digital, and Optable - launched the Ad Context Protocol, a framework built on Anthropic's MCP that aims to standardize how AI agents discover inventory, compare pricing, and activate campaigns across platforms. The launch divided the industry, with some welcoming it as necessary infrastructure and others warning it was premature. An industry veteran argued the same week that AdCP was "the tail wagging the dog," pushing automation tools before fixing fundamental structural issues in programmatic supply chains.

PubMatic has registered its MCP server with the IAB Tech Lab Agent Registry, which as of March 11, 2026 held 10 active entries all operating under the Model Context Protocol standard. PubMatic's entry carries a GPP and TCF ID - number 76 - linking its registry entry to its Global Privacy Protocol registration and connecting privacy compliance infrastructure to the agentic discovery layer. That detail matters for the AdRoll integration: both companies can point to a verified, auditable identity within the emerging governance layer that the IAB Tech Lab is constructing.

The security dimension deserves attention. Research published in July 2025 identified significant vulnerabilities in MCP implementations, with analysis suggesting "MCP security is completely broken" due to weaknesses in how the protocol handles tool interactions, particularly around tool poisoning attacks. Cross-platform agent communication of the kind AdRoll and PubMatic have demonstrated increases the surface area for such risks, since queries and responses traverse organizational boundaries. Neither company addressed security architecture specifically in the April 23 announcement.

AdRoll's investment in MCP as an open standard

According to the announcement, this collaboration builds on AdRoll's "early investment in MCP as an open standard for connecting AI systems to real marketing workflows." That framing positions MCP adoption not as a product feature but as a strategic infrastructure choice - one that allows AdRoll to connect with any platform that publishes an MCP-compliant interface, without bilateral integration work.

AdRoll unified its brand in August 2025, folding its B2B platform RollWorks into AdRoll ABM and launching an AI Assistant for campaign management. That restructuring positioned the company as an AI-powered multi-channel advertising platform rather than two separate products. The MCP integration with PubMatic extends that positioning into the infrastructure layer: AdRoll is not only building AI tools for its own users but also establishing the protocol-level connections that allow those tools to operate across the boundaries of its own platform.

PubMatic (NASDAQ: PUBM), which processed approximately 78 trillion impressions quarterly as of its second quarter 2025 financial results, has made agentic infrastructure central to its competitive strategy. An analysis of PubMatic's Q4 2025 results noted that the company is investing more aggressively than any comparable SSP in agentic AI positioning, even as its stock declined 51% over the prior year. The AdRoll partnership produces a concrete, named implementation to support that narrative.

What remains unsolved

The announcement is careful about what it does not claim. Nowhere does it suggest the integration is in general availability, nor does it specify what proportion of AdRoll's or PubMatic's customer base can currently access it. The PR outreach explicitly offered an interview with AdRoll's CEO to "walk through the MCP implementation, how the agents interact, and what still needs to be solved for broader adoption" - a construction that acknowledges open problems.

The advertising industry has documented persistent caution about autonomous agent authority over spending decisions. An industry analysis from January 2026 surveyed executives across agencies, platforms, and infrastructure providers, finding that while large language models are being integrated into workflow automation, platforms are deliberately preventing them from making autonomous bidding decisions. The AdRoll-PubMatic integration sits within that cautious zone: it surfaces root causes and recommended actions, but the record of whether and how quickly those actions are taken remains with human operators.

PubMatic's Optable partnership announced in March 2026 and Magnite's seller agent within SpringServe represent parallel developments in cross-platform agent workflows, all of which rely on standards adoption and governance infrastructure that is still being built. IAB Tech Lab's Programmatic Governance Council, announced April 21, 2026, lists PubMatic among its founding members alongside Omnicom, WPP, Dentsu, Disney, Amazon Ads, and The Trade Desk - an indication that governance frameworks are catching up to the technical experiments already underway.

The AdRoll-PubMatic MCP integration is a working example where most remain theoretical. It is narrow in scope - deal diagnostics, not full campaign management - and its production scale is not publicly specified. But it demonstrates something concrete: two independent platforms, with no shared ownership, can expose their internal diagnostics to each other's agents through an open protocol, and produce actionable output in a single coordinated workflow. That is what the industry has been debating in the abstract for the better part of a year.

Timeline

Summary

Who: AdRoll, an AI-powered multi-channel advertising platform based in San Francisco, and PubMatic (NASDAQ: PUBM), a supply-side advertising technology company, announced the integration. AdRoll CEO Vibhor Kapoor and PubMatic VP of Advertiser Solutions Alex Shephard were cited in the announcement.

What: An agent-to-agent integration using the Model Context Protocol, allowing AdRoll's demand-side AI agents to query PubMatic's supply-side diagnostic system directly to identify root causes of campaign delivery issues - including pacing constraints, creative blocks, and publisher-side factors - and surface recommended actions, within a single coordinated workflow.

When: The announcement was made on April 23, 2026. It validates PubMatic's agent-to-agent MCP spec, which was published in September 2025.

Where: The integration operates across AdRoll's demand-side platform and PubMatic's supply-side platform. Both companies are US-based. PubMatic is listed on the Nasdaq exchange.

Why: Deal troubleshooting in programmatic advertising has historically required manual investigation across multiple platforms, log exports, and bilateral support exchanges - a process measured in days. The MCP integration condenses that process into a single agentic workflow measured in minutes, addressing a structural inefficiency that affects advertisers running complex, deal-based buying strategies across independent platforms.

Share this article
The link has been copied!