Madhive today announced the expansion of its Maverick AI intelligence layer to include a suite of agentic capabilities embedded directly in its demand-side platform, alongside a new Model Context Protocol server that allows multiple large language models to interact with the platform through natural language commands.

Madhive today took its Maverick AI intelligence layer a step further, announcing the addition of agentic capabilities that will roll out across its local-first DSP over the coming year. The New York-based company, which describes itself as the only ad platform built specifically for local brands and agencies, made the announcement on May 27, 2026, framing the move as the next stage of an AI layer it first introduced in 2025.

The expansion covers the full campaign lifecycle. According to Madhive, the new agents handle everything from initial discovery and pitching through planning, activation, optimisation, and reporting - functioning as what the company calls an "operational thread" that runs continuously, including outside working hours.

What Maverick AI agents actually do

The agents are trained on more than a decade of proprietary data drawn from Madhive's local-first platform. According to Madhive, they ingest signals from over 50,000 daily campaigns. That data volume is significant: it means the models are exposed to the kind of fragmentation - geographic variance, budget variability, seasonality, and pacing irregularities - that defines local media buying at scale. Most general-purpose DSPs are not built around these dynamics, and Madhive's positioning rests on the claim that this specificity gives its agents a structural advantage over broader platforms.

Three core capabilities are highlighted. First, the agents accelerate media plan creation by drawing on historical performance and audience data, reducing the time required to move from brief to proposal. Second, they activate campaigns across local inventory with what Madhive describes as speed and precision. Third, they continuously optimise delivery and budget allocation in real time, rather than requiring manual intervention between reporting cycles.

"Local advertising has always been performance-driven at its core - defined by execution, accountability, and results," said Jim Wilson, CEO of Madhive. "With Maverick AI agents embedded directly into our DSP, we're enabling advertisers to close the loop between insight, activation, and performance in real time - unlocking a new level of efficiency and impact for local campaigns."

The language around "closing the loop" points at a persistent structural problem in local advertising. Media plans generated by one team are often activated and optimised by separate systems, with reporting arriving too late to influence in-flight decisions. Embedding agents at every stage - rather than adding them as a layer on top of existing workflows - is the technical bet Madhive is making here.

Alongside the agentic suite, Madhive has also launched a Model Context Protocol (MCP) server. This is a distinct product that operates at the infrastructure level, rather than within Madhive's own interface. The server allows multiple LLMs to connect directly to Madhive's platform, exposing functionality through a standard interface so that users can interact with campaigns, creatives, audiences, and publisher groups using natural language commands.

The MCP server is not designed solely for Madhive-native workflows. According to Madhive, it is intended to let enterprises bring local intelligence directly into their own tools, workflows, and systems - meaning a buyer could, in principle, query or manage campaigns through a third-party AI assistant or internal tool rather than only through Madhive's interface.

The adoption of MCP as an integration standard places Madhive in a broader industry movement. The advertising industry's engagement with MCP accelerated sharply throughout 2025, with Google releasing an open-source MCP server for its Ads API in October 2025, Amazon launching a closed beta for its own MCP Server in November 2025, and AdRoll and PubMatic building a cross-platform MCP integration in April 2026 to help AI diagnose programmatic deal failures. Madhive's server adds a local-media-specific node to that growing network of machine-readable advertising infrastructure.

The MCP standard, originally developed by Anthropic and later donated to the Linux Foundation, functions as a standardised connection framework between LLMs and external data sources. Its adoption across advertising platforms has been tracked extensively as the industry moves toward agentic workflows where AI systems can take actions across multiple platforms through consistent interfaces.

Performance claims and prior recognition

Madhive points to 2025 results as evidence that the underlying Maverick AI system already delivers measurable outcomes. According to Madhive, campaigns optimised by Maverick AI in 2025 saw a five-fold increase in ROAS and two to four times stronger ROI compared to baseline. The company also states that Maverick AI received AdExchanger's 2025 Award for Most Innovative CTV Technology.

These figures come from Madhive and have not been independently verified. The 5x ROAS figure is a performance claim made in reference to Maverick-optimised campaigns broadly, not a controlled study comparing identical campaigns run with and without the system. That context matters when evaluating the number, though its inclusion in an announcement directed at media buyers and agency partners signals how Madhive expects those claims to land commercially.

Fox's endorsement and what it signals

The announcement includes a statement from Michael Page, SVP of Digital Sales at Fox Television Stations, one of Madhive's named enterprise partners alongside Scripps and Hearst. "Insights have always been at the core of how FOX drives performance in local markets," Page said. "We're excited that Maverick AI agents will not only enhance that capability, they will fundamentally accelerate it, turning insight into immediate action and performance at scale."

The Fox endorsement matters for a specific reason. Fox Television Stations operates local broadcast properties across major US markets, making it a direct user of the local-scale DSP capabilities Madhive is expanding. An SVP-level statement focused on "immediate action" rather than planning or measurement suggests the velocity improvement - reducing the time between insight and execution - is the aspect of the product that Fox finds most compelling in practice.

Context: agentic AI and the DSP question

The launch arrives at a moment when agentic AI's relationship to traditional DSP infrastructure is actively contested. In July 2025, industry analyst Ari Paparo argued that autonomous AI systems could eventually automate the campaign setup, targeting, and optimisation functions that DSPs currently provide - potentially bypassing DSP infrastructure rather than augmenting it. Microsoft's closure of its Xandr DSP, announced with reference to "agentic" advertising futures, added weight to that argument.

Madhive's approach is the opposite of that scenario. Rather than treating agentic AI as a threat to DSP relevance, it is embedding agents inside its DSP - making the platform the place where agentic execution happens, not the layer being automated around. Yahoo DSP took a similar approach in January 2026, integrating agents that can autonomously monitor campaigns, diagnose problems, and execute corrective actions. Amazon launched its Ads Agent at its unBoxed 2025 conference. PubMatic's AgenticOS went live with active campaigns in January 2026.

What distinguishes Madhive's implementation is the local-media specificity. Most of these launches target national or performance-oriented advertisers. Madhive's focus on local fragmentation - the geographic patchwork of markets, audience sizes, and seasonal patterns that define local broadcast and CTV buying - is a narrower brief, and the claim is that a decade of data trained on exactly that brief produces agents that are qualitatively different from ones trained on national-scale campaigns.

A January 2026 survey by Proximic by Comscore found that 82 percent of advertising decision-makers now consider AI-powered optimisation essential when evaluating partners. That figure reflects conditions across programmatic advertising broadly, but it establishes the baseline expectation against which any DSP launch is now being judged. Madhive's agentic expansion is, in part, a response to that shift in buyer requirements.

The local SMB angle

Madhive also provides context on its broader reach through its Frequence subsidiary. According to Madhive, the combined platform powers campaigns for more than 50,000 small and medium-sized businesses every day. The challenge of giving SMBs access to CTV-grade campaign infrastructure has been a recurring theme across the ad tech industry in 2025 and 2026. Magnite's acquisition of streamr.ai in September 2025 targeted exactly that gap from the supply side. IAB Europe's Virtual Programmatic Day in April 2026 devoted significant discussion to the barriers keeping SMBs out of CTV - including minimum audience sizes, high creative costs, and execution complexity.

Madhive's argument is that its agents reduce the execution complexity barrier for exactly this segment. A local business that does not have a dedicated media team, or that relies on a broadcaster's sales operation rather than an independent agency, stands to benefit from agents that generate plans and optimise delivery without requiring specialist input at each stage.

Rollout timeline

Madhive has not announced a fixed date for full deployment of all agentic capabilities. According to Madhive, the suite will roll out over the next year. The MCP server appears to be available now, though the company has not specified access terms or pricing. The agents covering media plan creation, campaign activation, and real-time optimisation are the three functions named in the announcement, but the framing of an "operational thread" covering the full campaign lifecycle from discovery to reporting implies additional capabilities are planned.

The announcement is timed alongside the television Upfronts period, when national media commitments dominate industry conversation. Local media buyers are typically less visible during Upfronts, which may explain the positioning of this launch as addressing a market that has historically operated in the shadow of larger national commitments.

What this means for media buyers

For agencies and media professionals working with local broadcast or CTV inventory, the practical implication is that a major platform in this category is moving toward automated plan generation and continuous optimisation as defaults rather than optional features. That changes the workflow assumptions for campaign management, and potentially the staffing models of agencies and broadcaster sales teams who have relied on manual processes.

The MCP server introduces a second set of implications. If enterprises can connect their own LLMs and internal tools to Madhive's platform through a standardised interface, that shifts control over how campaigns are managed and queried away from Madhive's native UI and toward the buyer's own AI infrastructure. That is a meaningful change in how the buyer-platform relationship is structured - and one that mirrors similar moves in the broader programmatic ecosystem, where agentic AI standards debates continue to shape how platforms expose their capabilities to outside AI systems.

Timeline

Summary

Who: Madhive, a demand-side platform and AI solutions company headquartered at 225 Broadway, New York, built specifically for local brands, agencies, and media companies including Fox, Scripps, and Hearst.

What: The expansion of Maverick AI, Madhive's enterprise intelligence layer, to include a suite of agentic capabilities embedded across the full campaign lifecycle - from planning and activation through optimisation and reporting - alongside a new Model Context Protocol server allowing multiple LLMs to connect directly to the platform via natural language commands.

When: Announced on May 27, 2026. The full suite of agentic capabilities will roll out over the next year.

Where: The agents are embedded within Madhive's DSP platform, which operates across local media markets in the United States. The MCP server extends access to external enterprise tools and workflows.

Why: Madhive positions the launch as addressing the specific complexity of local media - geographic fragmentation, budget variability, seasonality, and pacing dynamics - that general-purpose DSPs are not built to handle. The agents are trained on more than a decade of proprietary local data and draw on signals from over 50,000 daily campaigns, with the stated goal of compressing the time between insight, activation, and performance measurement for local advertisers and the media teams that serve them.

Share this article
The link has been copied!