PubMatic announced the launch of PubMatic AgenticOS, an operating system designed to orchestrate autonomous, agent-to-agent advertising execution across premium digital environments. AgenticOS provides a system-level layer that allows agents to plan, transact, and optimise programmatic advertising with speed, consistency and control.
PubMatic is launching AgenticOS in close partnership with a group of advertisers, agencies, and publishers engaged in testing, shaping, and deploying agent-led workflows, providing real-world validation and innovation for the next phase of AI-native advertising.
With AgenticOS, advertisers define objectives, guardrails, brand-safety requirements, and creative parameters in their preferred LLM interface. PubMatic’s platform then operationalises that intent through a coordinated set of agents that plan, execute, and optimise campaigns continuously within defined guardrails, allowing teams to focus on strategy, creativity, and performance analysis rather than manual setup and troubleshooting.
From Agent Innovation to an Operating System
AgenticOS is a synthesis of PubMatic’s multi-year investment in agentic AI and interoperability. From the first agent-to-agent communication specification to yield optimisation, audience or inventory discovery and planning and deal troubleshooting agents, these innovations now come together in AgenticOS.
Unlike point execution layers or single-protocol connectors, AgenticOS runs on NVIDIA-accelerated computing and is embedded within PubMatic’s global infrastructure. This foundation enables large-scale, low-latency inferencing and coordination across millions of advertising transactions per second, supporting reliable execution in live markets where performance, scale, and accountability are essential.
In December 2025, PubMatic worked with independent agency Butler/Till and Geloso Beverage Group’s Clubtails to run an early agentic campaign on AgenticOS. Guided by natural-language input through Claude, PubMatic AgenticOS autonomously recommended campaign tactics, executed the buy, and optimized performance in real time within predefined parameters. This automation allowed the agency team to focus on higher-value strategic planning, creative development, and measurement, while the platform handled continuous execution and optimisation.
A Three-Layer Architecture Built for Agentic Intelligence
AgenticOS is built on PubMatic’s Architecture of Advertising Intelligence, a three-layer framework designed to support autonomous decision-making responsibly and at global scale.
The frameworks include, the infrastructure layer, NVIDIA-accelerated computing enables microsecond-level inference and real-time, privacy-safe data integration across tens of millions of auctions per second, delivering up to 5× faster decisioning, sub-millisecond response times, and significantly fewer auction timeouts.
The application layer, agentic capabilities embedded across PubMatic’s platform interpret intent expressed through protocols such as the Ad Context Protocol (AdCP) and Model Context Protocol (MCP). These agents automate and coordinate key advertising functions, including planning, forecasting, pacing, yield management, troubleshooting, and measurement, working together to continuously improve outcomes within defined constraints.
The transaction layer is where agentic intelligence becomes real-world execution. An interoperable execution environment connects agentic decisioning directly to PubMatic’s Activate buying platform, bringing together premium supply, in-flight campaign signals, and direct bidding infrastructure to transact in real time across Programmatic Guaranteed and Private Marketplace transactions. This layer ensures agent-led workflows remain compatible with existing buying models while enabling more intelligent, adaptive execution over time.
Accelerating Agentic Adoption Across the Ecosystem
To support real-world adoption, PubMatic is also launching the Agentic AI Acceleration Program, a focused initiative designed to help advertisers, agencies, publishers, and partners move from early testing to live agentic workflows within weeks. The program emphasises execution, governance, and measurable outcomes, supporting in-market deployments throughout Q1 2026.
PubMatic is inviting additional ecosystem partners to participate and help shape how agentic workflows are deployed, governed, and scaled across the open internet.
