Pacvue today launched an MCP server designed to connect its commerce media platform directly to enterprise AI tools including ChatGPT, Copilot, Gemini, and Claude, allowing brands and agencies to retrieve advertising reports from more than a dozen retail media networks through plain-language requests. The announcement, dated May 14, 2026, and distributed via GlobeNewswire from Los Angeles, marks Pacvue's first publicly available implementation of the Model Context Protocol - the open standard originally developed by Anthropic that defines how AI models communicate with external tools and data sources.

The first capability available through the server is Report MCP, which lets teams pull existing Pacvue advertising reports into the AI tool of their choice without manually navigating the platform, configuring export settings, or transferring files. The result is a formatted CSV or Excel file delivered inside the AI environment where the team is already working.

What the server actually does

The technical approach is straightforward. Pacvue has built its existing platform to be MCP-compatible, which means AI agents can connect to it natively using the protocol without any custom integration work on the advertiser's side. Report MCP sits on top of Pacvue's existing My Report engine, inheriting the same report types, permissions structures, and data access controls that teams already rely on.

According to Pacvue, the server connects to the full set of platforms available through that engine: Amazon, Walmart, Instacart, Kroger, Target, Sam's Club, Chewy, DoorDash, eBay, Bol, Mercado Libre, Criteo, and Citrus. New platforms and report fields added to Pacvue in future are picked up automatically - no separate configuration is required when coverage expands. The report types available include campaign performance, keywords and search terms, share of voice, inventory, and others that form the backbone of day-to-day retail media management.

Connection is supported from Claude, ChatGPT, Cursor, VS Code, and any other MCP-compatible client. Internal agent integrations using the same endpoint are also possible, meaning engineering teams can incorporate Pacvue data into broader automated workflows rather than being limited to consumer-facing AI tools.

"The way enterprise teams work is changing. AI tools are becoming a primary surface for analysis, decision-making, and execution, which means commerce media data needs to be accessible where those teams already operate," said Sunava Dutta, Chief Product Officer at Pacvue. "Pacvue's MCP server, based on the Model Context Protocol, creates a secure way for AI assistants and enterprise workflows to interact with commerce media reporting data. Report MCP is an important foundation for a broader roadmap that will expand how AI connects to commerce media intelligence and execution."

Dutta's framing of security as a distinguishing characteristic is worth noting. MCP implementations in advertising carry authentication and data access risks that have received industry attention. Security researchers identified tool poisoning vulnerabilities in MCP implementations in July 2025, highlighting how malicious actors could potentially manipulate how MCP servers communicate with client applications. Pacvue does not disclose the specific mechanisms used to secure the connection in today's announcement.

The operational problem it addresses

Enterprise advertising teams managing campaigns across a growing number of retail media networks face a consistent friction point: data lives in separate platforms, each with its own reporting interface, export process, and workflow requirements. For a team already using ChatGPT or Copilot for analysis and decision-making, pulling the necessary data has typically meant switching to Pacvue, locating the correct report, configuring it, exporting the file, and importing it back into whatever tool is being used for the actual work.

According to Pacvue, that friction is what Report MCP removes. Teams request data in plain language and receive it as a formatted report, delivered directly in the AI tool where they are already working. The plain-language interface removes the requirement to know which report type to select or which parameters to apply - the AI interprets the request and retrieves the appropriate output.

The company describes its platform as serving over 70,000 brands and agencies. As of 2025, according to Pacvue, it powers 12% of total retail media ad spend globally. Those numbers provide context for the scale at which the reporting friction exists - and the scale at which the integration could have operational impact, if adoption proves significant.

MCP's accelerating adoption in ad tech

Pacvue's launch arrives at a moment when the Model Context Protocol has moved rapidly from a niche developer standard to a foundational layer for advertising technology infrastructure. The trajectory is dense with milestones. Microsoft launched its Clarity MCP server in June 2025, enabling natural language queries for web analytics data. AppsFlyer introduced its MCP tool on July 17, 2025, targeting mobile measurement and attribution. Google Analytics released its MCP server on July 22, 2025. 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.

The retail media side of ad tech moved quickly as well. Amazon Ads launched a closed beta for its own MCP server on November 13, 2025, and entered open beta on February 2, 2026, at the IAB Annual Leadership Meeting - allowing external AI platforms including Claude, ChatGPT, and Gemini to interact with Amazon's advertising APIs through natural language. The Ad Context Protocol, built on Anthropic's MCP standard, launched in October 2025 to create a unified layer for campaign execution across platforms. Usercentrics acquired MCP Manager in January 2026 to extend privacy governance frameworks into AI-driven workflows.

Cross-platform agent-to-agent communication has also become a live use case. AdRoll and PubMatic announced an MCP-powered integration in April 2026 that allows AI agents to query deal diagnostics across the two platforms in real time - one of the first public implementations of bidirectional agent communication across the demand and supply sides of the programmatic stack. Meta opened its ad system to Claude and ChatGPT via MCP connectors on April 29, 2026, allowing AI agents to manage campaigns, catalogs, and ad signals directly.

Against this backdrop, 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. The registry represents an attempt to bring neutral, industry-managed discoverability to a rapidly expanding set of implementations that might otherwise fragment across vendor-specific documentation.

Pacvue's broader AI trajectory

Today's MCP launch is not Pacvue's first step into AI-driven workflow tooling. The company announced Pacvue Agent on April 14, 2026 - an AI agentic layer within its AI Outcome Engine that covers Amazon Ads at launch, with additional retailers planned throughout 2026. That product claims workflows up to 200x faster and time-to-insight up to 80x faster, converting AI recommendations into governed campaign execution steps. At the time of that announcement, MCP integration was described as in development. Today's launch represents the delivery of that commitment.

The product architecture distinguishes the two offerings. Pacvue Agent is embedded within Pacvue's own platform and Slack, operating as an agentic interface within the existing product. The MCP server, by contrast, is an outward-facing connection layer that allows external AI tools to reach into Pacvue rather than requiring users to go to Pacvue. Both address the same underlying problem - the gap between where analysis happens and where data lives - but from different directions.

In March 2026, Pacvue brought Reddit Ads into its Commerce Operating System, expanding the platform's reach beyond traditional retail media into community-driven discovery. In May 2026, Pacvue was named as a technology partner for ChatGPT ads alongside Adobe, Criteo, Kargo, and StackAdapt when OpenAI opened its self-serve Ads Manager to all US businesses. Each of these moves extends the platform's surface area across an advertising landscape that is restructuring rapidly around AI-native interfaces.

What Report MCP does not do

The announcement is precise about what has launched. Report MCP retrieves data. It does not execute campaigns, adjust budgets, modify bids, or take any action within Pacvue's platform. The capability is read-only in the sense that data comes out; instructions do not go in. Pacvue states that additional MCP capabilities are already in development, and that the roadmap will "expand how AI connects to commerce media intelligence and execution" - the word execution being the meaningful signal about where the product is headed.

The framing around permissions and data access is also deliberate. Because Report MCP is built on Pacvue's existing reporting infrastructure, the permissions that govern which data each user can access remain in place. An agency user with access to a specific client's data does not gain broader access by connecting through MCP. That architectural decision - inheriting rather than rebuilding the access control layer - reduces the risk that MCP connectivity inadvertently opens data to parties who should not have it.

The question of data quality is separate from the question of connectivity. Pacvue's existing reporting infrastructure has the same accuracy characteristics through MCP as it does through the platform's standard interface. The AI tool requesting the data does not improve or degrade the underlying numbers; it changes how they are accessed and in what format they are returned.

Competitive context

Pacvue is not the only retail media management platform moving toward MCP-based connectivity. The broader competitive dynamic runs in two directions. Platform-native agents - built by Amazon, Walmart, or the retail networks themselves - can offer depth within their own environments that third-party tools cannot easily replicate. Third-party tools like Pacvue offer breadth across retailers that no single platform can match. The MCP server reinforces that breadth argument: a single connection point providing access to data from thirteen retail media networks is a meaningful differentiator against a world where each network's own MCP server covers only that network's data.

The company's position is also shaped by the specific AI tools it supports. Connecting to Claude, ChatGPT, Copilot, and Gemini simultaneously means the integration works regardless of which AI platform an enterprise team has standardized on. That neutrality matters as enterprise AI stacks are still forming and vendor choices are not yet settled. A Pacvue MCP server that worked only with one AI tool would be significantly less useful than one that works across the major options currently in enterprise deployment.

Report MCP is generally available as of today's announcement, with no stated restrictions on availability or eligibility for existing Pacvue customers.

Timeline

Summary

Who: Pacvue, a Los Angeles-based commerce media platform founded in 2018, serving over 70,000 brands and agencies and powering approximately 12% of total retail media ad spend globally as of 2025. The announcement was made by Sunava Dutta, Chief Product Officer.

What: Pacvue launched an MCP server, with Report MCP as its first available capability. The server allows enterprise AI tools to retrieve Pacvue advertising reports from more than a dozen retail media networks - including Amazon, Walmart, Instacart, Kroger, Target, Sam's Club, Chewy, DoorDash, eBay, Bol, Mercado Libre, Criteo, and Citrus - through plain-language requests, returning results as formatted CSV or Excel files. Connection is supported from Claude, ChatGPT, Cursor, VS Code, and any MCP-compatible client.

When: The announcement was made on May 14, 2026, via GlobeNewswire. Report MCP is generally available as of that date.

Where: Pacvue is headquartered in Los Angeles, with offices in Seattle, Chicago, New York, Washington D.C., London, Shanghai, and Tokyo. The MCP server operates through Pacvue's existing My Report engine infrastructure.

Why: Enterprise advertising teams increasingly use AI tools for analysis and decision-making, but the reporting data that feeds those workflows still requires navigating separate platforms, configuring reports, and manually exporting files. Pacvue's MCP server is designed to make commerce media data reachable wherever those workflows happen, without rebuilding existing permission structures or requiring separate access configuration.

Share this article
The link has been copied!