ARTF Glossary

This glossary summarizes key terms used throughout the ARTF.ai site. Definitions are aligned with the Agentic RTB Framework (ARTF) and related execution concepts.

Agentic RTB Framework (ARTF)

A technical framework defined by the IAB Tech Lab that specifies a containerized execution model for real-time bidding. It standardizes how host platforms run and orchestrate agent services that can enrich, evaluate, or modify the bidstream.

Host platform (orchestrator)

The supply-side platform, demand-side platform, exchange, ad server, or curation platform that operates the auction or decisioning workflow. The host controls the infrastructure, decides which agents to invoke, governs which data is shared, and chooses which mutations to apply.

Agent service (agent)

A third-party or first-party service packaged as a container. Examples include identity resolution, fraud detection, contextual or audience enrichment, bidding logic, and measurement. Agents are invoked by the host to perform specific tasks on the bidstream.

Container

A standardized, lightweight, and portable software package that bundles an agent's code and dependencies. Containers run inside the host platform's environment, providing isolation and resource control while enabling low-latency execution.

Protected bidstream mutation

The ARTF model in which agents do not receive unrestricted access to the entire bid request. Instead, the host provides only the data required for a given task and accepts proposed changes as structured mutations, applying or rejecting them under its own policies.

OpenRTB Patch

A structured mechanism for representing changes to an OpenRTB object as a set of atomic mutations. Each mutation specifies an operation (such as add, replace, or remove), a target path in the OpenRTB object, and an associated intent. Hosts use OpenRTB Patch to integrate agents' proposed changes into the live bidstream.

Intent

A declarative label that describes the purpose of a mutation or of an agent's capability (for example, identity resolution, fraud detection, deal activation, or bid adjustment). Intents help hosts understand, order, and govern the behavior of agents.

True bidding agent

A containerized agent controlled by an advertiser or agency that implements proprietary bidding logic or models. When deployed into an ARTF-compliant host environment, such an agent can score impressions and make bidding decisions according to the buyer's own rules.

"Package once, deploy broadly"

A deployment model enabled by ARTF in which technology partners package their logic into a single standardized container image and manifest. Any ARTF-compliant host can adopt this image, reducing the need for bespoke, per-partner integrations.

Model Context Protocol (MCP)

A JSON-RPC-based protocol designed to allow AI models and tools to access external resources and services. In the context of ARTF, MCP is referenced as a complementary mechanism for model-to-service or model-to-agent interactions.

Agent-to-Agent (A2A) communication

A pattern in which agents collaborate or exchange information with each other, in addition to interacting with the host. ARTF anticipates A2A scenarios as agentic systems mature, subject to host governance and security constraints.