ARTF.ai

Independent reference for the Agentic RTB Framework (ARTF) for advertisers, agencies, and ad tech

ARTF is a framework defined by the IAB Tech Lab that specifies a containerized execution model for real-time bidding. It enables host platforms to run third-party "agent services" as co-located containers within their own infrastructure, using a standardized API to enrich and modify the bidstream during the auction.

To explore ARTF in more depth:

Why ARTF was introduced

The framework is a response to structural limits of the legacy programmatic architecture:

  • High latency. Bid requests often traverse multiple external services over HTTP, leading to end-to-end latencies in the range of hundreds of milliseconds per impression.
  • Operational complexity. Each new vendor typically requires a bespoke integration, creating a fragile "bolt-on" ecosystem of external calls.
  • Data exposure. Passing full or partial bid requests to multiple external endpoints increases the risk of data leakage and complicates compliance.

ARTF addresses these issues by moving execution into a controlled, co-located environment operated by the host platform.

What ARTF changes

  • Co-located execution. Third-party services are packaged as lightweight containers and run within the host platform's data center or cloud environment.
  • Standardized API and patch model. Hosts interact with agents via a defined API that supports "protected bidstream mutation" through OpenRTB Patch and declarative "intents".
  • Latency reduction. By eliminating most external network hops, ARTF is designed to reduce bid request/response times by up to approximately 80%, from typical ranges of several hundred milliseconds to around 100 ms, depending on implementation.
  • Controlled data sharing. Hosts determine which data elements are exposed to each agent and which mutations are applied, maintaining control of data and service-level agreements.
  • Foundation for agentic AI. The framework provides the execution layer for systematic agent integration today and anticipates future autonomic agents via protocols such as the Model Context Protocol (MCP) and agent-to-agent (A2A) communication.

Vision

ARTF is intended to provide a durable execution layer for a more efficient and intelligent programmatic ecosystem:

  • High-performance foundation. Replace high-latency, bolt-on integrations with standardized, co-located services that make better use of the auction window.
  • Secure collaboration on valuable data. Enable hosts and data owners to work with sensitive signals in real time while keeping control of where data flows and how it is used.
  • Path to agentic and AI-native trading. Prepare the infrastructure for systematic agents today and for future autonomic agents that can participate directly in real-time bidding under host governance.

Who this site is for

This site summarizes ARTF for practitioners in:

  • Advertisers: brands and in-house teams evaluating how to run proprietary bidding logic closer to the impression and gain finer control over optimization.
  • Agencies: organizations buying on behalf of many advertisers and coordinating strategies across platforms.
  • Trading desks: teams responsible for cross-campaign and cross-client execution, often operating alongside agencies.
  • Host platforms: SSPs, ad exchanges, publisher ad servers, retail media supply platforms, and buy-side hosts (DSPs and other buying platforms) considering support for containerized agents and standardized orchestration.
  • Intelligence and infrastructure partners: identity, fraud, enrichment, measurement, and other technology partners assessing a "package once, deploy broadly" model.
  • Media and data owners: publishers, broadcasters, apps, retail media operators, and other first-party data holders seeking secure, real-time collaboration using sensitive or high-value data.

This site consolidates key information about ARTF and its practical implications for the programmatic advertising ecosystem. Key terms are summarized in the ARTF glossary.