By The ad tech industry's agentic AI push just got its first standards-compliant testing environment, and with it, a clearer picture of what "agentic media buying" actually means in practice.
IAB Tech Lab and Kochava announced on March 24 that Kochava has integrated the AAMP (Agentic Advertising Management Protocols) Buyer Agent SDK as an open-source workspace inside its StationOne platform. StationOne exits closed beta with this launch, making the workspace available to anyone who downloads the app.
This matters because until now, agentic advertising has been a collection of competing vendor demos and aspirational blog posts. AAMP gives it a shared spec. StationOne gives it somewhere to run.
What AAMP actually is
AAMP is IAB Tech Lab's umbrella initiative for standardizing how AI agents operate inside advertising systems. The organization formally named it on February 26, 2026, consolidating earlier agentic work under three pillars:
- Agent Foundations covers execution infrastructure, including how agents interact with real-time bidding environments. This is where ARTF, the real-time framework spec, lives.
- Agentic Protocols defines the management layer: how buyer and seller agents discover inventory, negotiate deals, and complete setup tasks. It builds on existing standards like OpenDirect, AdCOM, and OpenRTB rather than replacing them.
- Trust and Transparency provides the Agent Registry, a free directory where anyone can register and verify agents operating in the ecosystem.
The critical design choice: AAMP extends existing IAB Tech Lab standards rather than starting from scratch. OpenDirect handles direct transactions. AdCOM handles object models. OpenRTB handles bidding. The agentic layer sits on top, using Model Context Protocol (MCP) as the connective tissue between AI models and these established specs.
"AAMP is built to support AI-driven workflows on open, trusted standards. Making this workspace open-source within StationOne opens the door for broader industry participation and continued development," said Anthony Katsur, CEO of IAB Tech Lab.
What StationOne does
StationOne is a desktop application, not a cloud service. Kochava CEO Charles Manning described it to Digiday as a "desktop-based orchestration layer" that connects AI models and ad tech tools via API keys and MCP integrations.
Manning called it an "integrative AI hub" that brings together "various different tools with various different skills across various different AI models." Think of it as a Slack-like interface where you plug in your own model APIs, connect them via MCP, and orchestrate workflows across DSPs, data tools, and measurement platforms.
The IAB Tech Lab workspace inside StationOne exposes 19 specialized skills across 8 functional areas, according to PPC Land's detailed breakdown. Those areas map to the full campaign lifecycle:
- Foundation and Setup establishes organizational structures, account hierarchies, and tiered pricing
- Planning and Strategy builds campaign briefs optimized for agentic execution through the OpenDirect protocol
- Inventory Discovery uses natural language queries to find matching inventory across publishers
- Deal Structuring handles Programmatic Guaranteed, Preferred Deals, and Private Auctions
- Campaign Execution translates briefs into fully executed OpenDirect bookings
- Creative Management handles asset registration, approval workflows, and rotation weights
- DSP Activation extracts Deal IDs and provides platform-specific steps for The Trade Desk, DV360, and Amazon DSP
- Performance and Optimization monitors delivery pacing and recommends mid-flight adjustments
All 19 skills run through the IAB Tech Lab's reference implementation MCP Server. The source code for skills, agent definitions, and system prompts is visible to anyone in the workspace.
Why the timing matters
As Digiday noted, headcount reductions across ad tech are forcing campaign teams to do more with less. That pressure has triggered a rush toward agentic tools, which has created its own fragmentation problem: multiple competing approaches are already emerging, from AdCP to various vendor-specific frameworks, each with its own logic and taxonomy.
That fragmentation risk is real. The industry has seen this movie before with identity solutions, header bidding wrappers, and attribution models. Without early convergence on interoperability, agentic tools risk repeating those same balkanization patterns but now compounded by LLM-driven workflows that each speak a different language.
Manning told Digiday that Kochava is "talking to agencies about how this can help augment their agentic strategy," noting that StationOne can help bring consistency for campaign teams juggling multiple platforms with disparate workflows.
What this changes for you Monday morning
Honestly, not much yet. This is a testing environment, not production infrastructure. You cannot run real campaigns through it today.
But here is what you can do: download StationOne, open the IAB Tech Lab workspace, and see exactly what "agentic media buying" looks like when it follows a shared standard. Inspect the 19 skills. Read the system prompts. Understand how buyer and seller agents are supposed to negotiate using OpenDirect primitives. That visibility is the whole point of making it open source.
If you run campaigns through DSPs, the DSP Activation skill is the one to watch. It currently provides activation steps for The Trade Desk, DV360, and Amazon DSP. How well those handoffs work in practice, once the system moves from sandbox to live campaigns, will determine whether AAMP becomes a real workflow improvement or another spec that lives on a GitHub page.
The concrete next step: join the Agentic Task Force at IAB Tech Lab if you want to shape the requirements. If you just want to understand the technology, install StationOne and break things in the sandbox. Both options beat waiting for vendors to tell you what agentic means.
Devon Park covers ad tech and measurement for The Daily Vibe.
