On March 1, 2026, IAB Tech Lab launched the Agent Registry — a directory where buyer and seller agents in programmatic advertising can discover each other. By March 11, it had ten entries. PubMatic. Amazon. Dstillery. Optable. All running on Model Context Protocol. The registry itself is an MCP server — agents discovering agents.
D/Cipher is not in it.
What Got Built While Nobody Was Looking
In January 2026, IAB Tech Lab published AAMP — Agentic Advertising Management Protocols. Three pillars: ARTF for real-time execution (80% latency reduction over current OpenRTB), open-source agent schemas for buyer/seller interoperability, and the Agent Registry for trust and discovery. AAMP doesn't replace the programmatic stack. It extends it for machine-speed operation.
Here's what happened next.
Nine milestones in five months. Not one involves D/Cipher. Not one involves People Inc.
What PubMatic Already Proved
At an IAB Europe webinar in April, PubMatic demonstrated what agent-to-agent advertising actually looks like. Their MCP server analyzed 1,200 incoming bids and 2,000 winning bids, identified a 1% DSP bid adherence rate as the root cause of poor delivery, and surfaced the fix — in seconds. What used to take days of support tickets between a DSP and an SSP now takes a single agent query.
By end of Q1, PubMatic had run 30+ fully autonomous end-to-end agentic campaigns and over 1,000 direct publisher deals through AgenticOS. The AdRoll integration proved the cross-platform case: a DSP agent querying an SSP's MCP server for deal diagnostics — pacing issues, creative blocks, publisher-side constraints — without a human opening a ticket.
This is supply-side infrastructure. PubMatic is building what SSPs look like when the buyer is a machine. D/Cipher should be building what publisher intelligence looks like when the buyer is a machine.
The Interface D/Cipher Doesn't Have
D/Cipher has two layers, as I covered in The Signal Migration. Layer 1 is a pre-computed taxonomy — every page across 40+ brands tagged with content signals, audience intent, and contextual categories. Layer 2 is the real-time session layer — what a specific user is reading right now, computed on the fly using OpenAI embeddings.
Layer 1 is exactly what an MCP server would expose. Contextual signals, content taxonomy, intent categories — structured data that a buyer agent could query before deciding where to place an ad. A media buyer's agent could ask: show me all available inventory tagged with 'wedding planning' intent, cross-referenced with 'luxury spending' signals, on pages with brand-safety scores above 90. D/Cipher could answer that query better than anyone in publishing.
But it can't answer the query at all, because nobody can ask it. There is no MCP server. There is no entry in the Agent Registry. There is no AAMP integration. D/Cipher's contextual intelligence lives behind a sales team, accessed through a managed service or a PMP deal.
Who Is at the Table
Building Agentic Infrastructure
- PubMatic — AgenticOS, MCP server, Agent Registry
- Magnite — Seller Agent on SpringServe
- Amazon — Agent Registry participant
- Disney — Magnite expansion partner
- The Trade Desk — Governance Council
- Kochava — StationOne AAMP workspace
- Omnicom, WPP, Dentsu — Governance Council
Not at the Table
- D/Cipher — not in Agent Registry
- D/Cipher — not in Governance Council
- D/Cipher — no MCP server
- D/Cipher — no AAMP integration
- D/Cipher — not in Magnite CTV tests
The best contextual intelligence in publishing, invisible to every buyer agent being built.
The Managed Service Trap
D/Cipher currently operates as a managed service. An advertiser works with People Inc's sales team. The team sets up targeting using D/Cipher's taxonomy. The campaign runs through PMPs. Performance is reported. It works — D/Cipher campaigns claim 2x the performance of cookie-based targeting.
But managed services don't scale in an agent-to-agent world. When Omnicom's AdCP is sending buyer agents to negotiate directly with seller agents, when PMG's Alli is executing campaigns through AAMP protocols, when PubMatic's MCP server is fielding diagnostic queries from AdRoll's agents in real time — the question isn't whether D/Cipher's signals are good. The question is whether a buyer agent can find them.
Consider the progression:
2024: Buyers call the sales team. D/Cipher is a differentiator.
2025: Buyers set up PMPs. D/Cipher is still accessible, still human-mediated.
2026: Buyer agents query seller infrastructure. D/Cipher is invisible.
2027: If buyer agents can't find your signals, your signals don't exist.
This isn't hypothetical. PubMatic has already run a thousand deals through agentic infrastructure. Magnite is testing with Disney and Samsung. The Kochava StationOne beta has 19 buyer-agent skills running on IAB Tech Lab's reference MCP server. The infrastructure is being built now, and whoever isn't in the registry when buyer agents go looking won't be in the consideration set.
What D/Cipher Should Build
1. An MCP server exposing contextual signals. D/Cipher's Layer 1 taxonomy — content categories, intent signals, brand-safety scores, audience crossover data — packaged as MCP tools that buyer agents can query. This is the minimum viable interface to the agentic ecosystem. PubMatic built theirs. It works. The pattern exists.
2. An Agent Registry entry. Register with IAB Tech Lab. Show up where buyer agents look. All ten current entries use MCP. PubMatic is the only one with a GPP/TCF privacy ID — D/Cipher's first-party, cookieless foundation would be a natural fit for privacy-compliant agent discovery.
3. AAMP seller-agent capability. Build a seller agent that can negotiate inventory access, communicate available taxonomy segments, and provide real-time campaign diagnostics — the same way PubMatic and Magnite already do on the supply side.
None of this requires abandoning the managed service model. It requires building a parallel machine interface alongside the human one. The managed service continues for complex, high-touch campaigns. The agentic interface handles the growing volume of machine-mediated buying.
The Timing Problem
People Inc's management attention is elsewhere. The $18 billion MGM bid landed June 1. That bid will consume executive bandwidth for months — board negotiations, financing commitments, regulatory filings, gaming commission approvals across jurisdictions.
Meanwhile, the ad tech industry is standardizing agent-to-agent protocols. Standards congeal fast. The Programmatic Governance Council — Omnicom, WPP, Dentsu, Disney, Amazon, The Trade Desk, PubMatic — is writing the rules. D/Cipher's absence means those rules will be written without the publisher perspective that D/Cipher uniquely represents.
IAB Tech Lab's CEO warned of "several false starts" and years of experimentation ahead. That's true. But standards set during false starts have a way of surviving — OpenRTB was messy in 2010, and it still runs programmatic today. Whoever shapes AAMP shapes the market it governs.
D/Cipher built the best contextual intelligence in publishing. The question is whether anyone — or anything — will be able to find it.