IAB Agentic Standards.
A practical guide to AAMP, ARTF, Agentic Audiences, Agent Registry, and the IAB Tech Lab roadmap for agentic advertising.
IAB Tech Lab's agentic work is not a single spec. It is an umbrella initiative for bringing agentic execution into the advertising ecosystem through foundations, protocols, and trust infrastructure built on existing industry standards.
Agentic advertising needs more than agents. It needs standards for execution, signal exchange, identity, governance, transparency, and audit.
- AAMP v1.0
- ARTF v1.0 public comment
- Agentic Audiences released
- Agent Registry live · Mar 2026
- Agentic Bid / Mobile Q2 2026
Fast read
- What it is
- AAMP is IAB Tech Lab's umbrella initiative for agentic advertising standards.
- Core pieces
- ARTF for containerized real-time execution, Agentic Audiences for signal exchange, Agent Registry for transparency, and agentic versions of existing standards.
- What it is not
- Not a single replacement for OpenRTB, AdCOM, OpenDirect, or Deals API.
- Why it matters
- Agentic advertising must extend proven transaction standards instead of creating disconnected automation layers.
- Best for
- Adtech, publishers, platforms, agencies, measurement companies, data vendors, and product teams building agentic advertising infrastructure.
- Watch-out
- The standards are evolving quickly. Validate official docs, repos, and public-comment status before implementation.
What AAMP is.
AAMP — Agentic Advertising Management Protocols — is IAB Tech Lab's umbrella initiative for agentic advertising work. It organizes the agentic roadmap across foundations, protocols, and trust infrastructure.
| Layer | Job | Examples |
|---|---|---|
| Agentic Foundations | High-performance execution and control | ARTF, container runtime, real-time services |
| Agentic Protocols | Schemas, tools, and open-source reference implementations | Agentic Direct, Agentic Deals, Agentic Audiences, Buyer / Seller SDKs |
| Trust & Transparency | Identity, disclosure, accountability | Agent Registry, verification, audit |
The three pillars of AAMP.
- 01
Agentic Foundations
The execution layer: how agent services can operate within host platforms, including real-time environments.
- 02
Agentic Protocols
The management and workflow layer: how buyer and seller agents discover, negotiate, transact, exchange signals, and complete setup tasks.
- 03
Trust & Transparency
The accountability layer: how the ecosystem verifies agents, understands authority, and preserves trust.
ARTF: the real-time container runtime.
ARTF — Agentic Real Time Framework — defines a foundation for agent services operating inside a host platform. The model uses containers deployed into host infrastructure so the orchestrating platform can call agent services directly while preserving latency, cost, privacy, and operational controls.
- Containerized (OCI-compliant) service agents deployed into host infrastructure
- Direct calls by the orchestrating / host platform
- Bidstream mutation via declared intents the host can accept or reject
- Use cases: identity resolution, segmentation / deal activation, fraud detection, bid valuation
- Host platforms retain data control and SLAs; agents run with restricted network access
- Written primarily for engineers and product managers implementing the framework
Status: released as v1.0 for public comment (the standards page indicates finalization is forthcoming, with v2.0 in development). The press release describes reducing bid request/response time up to 80% — a design target on the page, not a guaranteed benchmark.
Agentic Audiences: signal exchange for agents.
Agentic Audiences, formerly UCP / User Context Protocol, defines how intelligent agents exchange identity, contextual, and reinforcement signals. The important shift is the use of embeddings as a compact, privacy-preserving, interoperable representation for signal exchange — the same operating idea as signal containerization.
- Formerly the User Context Protocol (UCP)
- Donated by LiveRamp to IAB Tech Lab
- Three signal types: identity, contextual, and reinforcement
- Encoded as dense embeddings of 256–1024 dimensions
- Designed to support sub-100ms response times for real-time bidding (a design target)
- Open-source repository (specification under CC BY 4.0; implementations under Apache 2.0)
Signals travel as compact vectors — typically 256–1024 dimensions — instead of raw identifiers:
{
"signal_type": "contextual",
"embedding": {
"dim": 512,
"vector": [0.0123, -0.0456, 0.0789, "… 512 floats …"]
},
"ttl_seconds": 1800,
"privacy": { "raw_pii": false }
}Agent Registry: trust and transparency.
Agentic workflows require participants to know which agents exist, who operates them, what they are authorized to do, and how they should be verified. Agent Registry is the trust and transparency layer in the IAB Tech Lab roadmap.
- Agent identity, verification, and disclosure
- Registrants associate a GPP ID (including TCF GVL IDs) for validation
- A browsable directory plus a REST API; the registry is itself an MCP server
- Went live March 3, 2026; free to register; described as a "test and learn" phase
- Seller-agent verification ("ads.txt for agents") is a stated roadmap item, not a current feature
Built on existing advertising standards.
The IAB Tech Lab approach starts from existing standards because those standards encode years of transaction, object, taxonomy, privacy, and measurement knowledge. Agentic systems need to reference those standards, not bypass them.
OpenRTB
Real-time bidding objects; now enabled to carry Agentic Audiences.
AdCOM
The common object model — basis for Agentic Ad Objects (now MCP-capable).
OpenDirect
Direct-deal workflow — basis for Agentic Direct (now MCP-capable).
Deals API
Programmatic deals — basis for Agentic Deals (now MCP-capable).
Taxonomies
Content, Audience, Ad Product, and Privacy taxonomies agents read.
Agentic Direct / Deals / Ad Objects
Existing standards made MCP-capable for agents.
Buyer / Seller Agent SDKs
Open-source reference agents built on AdCOM, OpenDirect, Deals API, OpenRTB.
Agentic Bid / Agentic Mobile
Stated as MCP-capable in Q2 2026 — track official status.
AdCP and AAMP: not a simple either / or.
The useful question is not which acronym wins. It is which layer solves which problem. AdCP is a task and workflow language for advertising agents. AAMP is IAB Tech Lab's umbrella for extending existing standards into the agentic era. ARTF is the execution runtime; Agentic Audiences is a signal exchange layer. These can be complementary when boundaries are clear.
AdCP 3.1 RC
RoleAgent task and workflow protocol
Best questionWhat can the agent discover, ask, buy, activate, govern, or report?
Main risk if misreadTreated as complete infrastructure rather than a workflow layer
AAMP v1.0
RoleUmbrella standards initiative
Best questionHow does agentic advertising extend existing standards?
Main risk if misreadTreated as one monolithic spec
ARTF public comment
RoleContainerized real-time execution
Best questionHow can services run inside host platforms with low latency and control?
Main risk if misreadConfused with a buyer workflow protocol
Agentic Audiences released
RoleEmbedding-based signal exchange
Best questionHow do agents exchange identity, contextual, and reinforcement signals?
Main risk if misreadTreated as another segment taxonomy
OpenRTB / AdCOM / Deals API / OpenDirect foundational
RoleFoundational transaction and object models
Best questionWhat existing market objects should agentic systems reference?
Main risk if misreadBypassed by new agent workflows
Implementation lens by company type.
Pick your company type to see what to build first — and the AAMP frameworks and standards it leans on.
ARTF containers, bidstream mutation, deals, seller agents, agent registry, signal exchange.
Buyer agents, agentic deals, signals, governance, OpenRTB / AdCOM alignment, measurement.
Seller agent, inventory discovery, OpenDirect, deal workflow, content monetization, agent registry.
Agentic Audiences, embeddings, identity / context / reinforcement signals, provenance, activation.
Buyer agents, governance, direct and programmatic workflows, reporting, approval, audit.
Event flow, audit, outcome reporting, fraud, suitability, trust.
Agentic ad objects, creative formats, approvals, brand safety, content monetization.
Standards watchlist.
What to track as these specifications evolve. Roadmap timing reflects official statements at the validation date and can change.
AAMP roadmap
Track new releases and reference implementations.
ARTF v1 / v2
Track finalization status, sample code, and container runtime requirements.
Agentic Audiences
Track embedding schemas, signal types, privacy guidance, and repository changes.
Agent Registry
Track registration requirements, verification model, and adoption.
Agentic Direct / Deals / Ad Objects
Track how existing standards become MCP-capable.
AdCP 3.1 RC
Track compatibility, task changes, and implementation guidance.
OpenRTB / AdCOM bridge
Track how real-time buying objects connect to agentic workflows.
Governance and audit
Track human-in-the-loop, signature, consent, and accountability requirements.
Scale comes from standards, not better prompts.
Agentic advertising will not scale because agents get better at writing prompts. It will scale when agents can operate inside standards-based systems with clear schemas, runtimes, permissions, registries, provenance, audit, and outcome measurement.
- Use existing standards wherever possible.
- Define the layer before debating the acronym.
- Do not let agentic workflows bypass accountability.
- Treat embeddings as signal infrastructure, not magic.
- Treat containers as execution infrastructure, not generic AI wrappers.
- Use protocols to reduce ambiguity, not to create theatre.
Where this fits in the full standards stack.
The layer map
AAMP, ARTF, and Agentic Audiences define the agentic runtime, signal-exchange, and trust layer. Around it sit the transaction rails (OpenRTB, AdCOM, OpenDirect, Deals API), privacy and consent constraints (GPP, TCF, platform and OS rules), measurement trust (verification, OM SDK, accreditation), and the research evidence that validates outcomes.
Building against the agentic standards layer?
Map the workflow, standards layer, runtime, signal model, governance path, and business outcome before choosing what to implement first.