A seller agent is software that represents a publisher's inventory to AI buying agents. It lets an authorized buying agent find suitable inventory, check it against a campaign's needs and start a deal, all through a standard machine-readable interface. For publishers, the real change is not that machines buy media, but that the first conversation about your inventory may now happen between software, so how your inventory is described and represented matters more than ever.
What does a seller agent actually do?
Think about how a buyer finds publisher inventory today. A planner sends a brief, a sales or representation team replies with options, there is some back and forth on formats, audiences and price, and eventually a deal is set up. A seller agent takes the discovery and matching part of that exchange and makes it available to software.
In practical terms, a seller agent usually helps with four jobs:
- Describing inventory in a structured way that a machine can read, such as site or app, content category, geography, device, format and size.
- Answering questions from buying agents about what is available and on what terms.
- Matching inventory to a buyer's stated goals.
- Starting a transaction, such as creating a deal that the buyer can then activate.
The buying side is handled by a buyer agent, which works for an advertiser or agency. The seller agent works for the sell side. That split matters, because it means the seller agent reflects the publisher's and representation partner's rules, not the buyer's wishes.
Why are seller agents appearing now?
Two things have come together. First, AI applications have become far better at following goals across several steps. Second, open standards now exist for connecting those applications to outside systems.
The most widely discussed is the Model Context Protocol, or MCP. Its own documentation describes MCP as "an open-source standard for connecting AI applications to external systems", and compares it to a USB-C port for AI applications (modelcontextprotocol.io). The MCP specification describes servers that offer tools (functions an AI model can call) and resources (data the model or user can read). A seller agent built to be MCP-compatible can therefore describe what it offers in a way that compatible buying agents already understand, without every buyer building a custom connection. You can read a short definition of the protocol on our MCP glossary page.
How could seller agents change publisher representation?
Publisher representation has always been about translation. A representation partner learns what a publisher's audience, content and formats are really worth, then explains that to buyers in the terms they plan with. Seller agents do not remove that job. They change who is listening.
Inventory descriptions become a sales asset
When a human buyer reads a media kit, they can fill gaps with a phone call. A buying agent works from the data it is given. If your CTV inventory is labelled vaguely, or your app placements are missing format details, an agent may simply never surface them. Clean, honest, structured inventory data could become one of the most valuable things a representation partner maintains.
Discovery could reach buyers you would not have pitched
A small, high-quality niche publisher rarely makes the shortlist of a large agency's planners, because there is not time to evaluate hundreds of small sites. A buying agent can evaluate many more options against a brief. Where the inventory is well described and represented, a seller agent could help niche and mid-size publishers be considered on their merits.
The pace of the first conversation changes
Agents can ask and answer in seconds. That does not mean deals close in seconds, because pricing, quality checks and approval still apply. It does mean the early stage of matching inventory to a brief could become much faster.
Example (illustrative)
A mid-size regional news publisher has display, native and outstream video placements on desktop and mobile web, with strong local audiences in two countries. Its representation partner describes that inventory by IAB category, geography, device, format and size, with the deal types it supports. A buying agent working on a brief for a local services advertiser asks the seller agent for news inventory in those countries, on mobile, with video. The seller agent returns the matching placements and terms. The buying agent requests a deal, receives a Deal ID and passes it to the advertiser's buying platform for activation. The publisher's team never had to answer a cold email, yet every option shown was one they had approved.
What stays in the publisher's control?
This is the question most publishers ask first, and it is the right one.
A seller agent is a sell-side tool. It can only offer what it has been set up to offer. Pricing, deal types, which inventory is included and which buyers have access all sit with the publisher and its representation partner. The MCP specification itself puts user consent and control at the centre, stating that users must retain control over what data is shared and what actions are taken (MCP specification).
Activation still runs through familiar plumbing. A Deal ID is the same reference buyers already use for private marketplace and programmatic guaranteed deals. The agent helps create the agreement; delivery, measurement and billing follow the normal path.
What does a seller agent not do?
It is worth being clear about limits, because the topic attracts a lot of hype.
- It does not replace your ad server, header bidding setup or SSP partners.
- It does not make quality or brand safety checks unnecessary.
- It does not decide your pricing strategy for you.
- It does not open your inventory to anyone who asks, at least not in a well-run setup.
| Question | Human-led representation | With a seller agent |
|---|---|---|
| Who finds the inventory? | Planners and sales teams | Authorized buying agents, plus people |
| What do they read? | Media kits, emails, calls | Structured inventory data |
| Who sets terms? | Publisher and representation partner | Still the publisher and representation partner |
| How is a deal activated? | Deal ID in a buying platform | Deal ID in a buying platform |
How IncrementX approaches the seller agent
IncrementX is a global media representation partner first, and we see a seller agent as a new route for represented inventory to reach demand, not as a replacement for representation.
IncrementX Seller Agent is an MCP-compatible interface. Authorized AI buying agents can use it to discover IncrementX inventory by site or app, IAB category, geography, device and environment, ad format, ad size, inventory type, pricing, deal type and availability. They can evaluate and match that inventory to their needs, initiate deal creation and receive Deal IDs used to activate through supported buying platforms. Access is by request. It is not a public tool.
That sits alongside the rest of our work. Our team still does the inventory understanding, demand mapping and packaging described on our media representation page, and that groundwork is what makes inventory worth discovering in the first place. Seller Agent is designed to carry that work to a new kind of buyer. The wider agent-to-agent buying workflows around it are being built with partners, and we describe that side in more detail in our guide to buyer agents and seller agents.
Questions to ask before your inventory goes agent-ready
If you are a publisher thinking about this shift, a few practical checks go a long way:
- Is every placement described with category, geography, device, format and size?
- Are your deal options (open auction, PMP, PG, curated) clearly defined?
- Does your representation partner control who can access your inventory through any agent interface?
- Are your ads.txt and sellers.json entries accurate, so buyers can verify the path to your inventory?
- Do you know how a Deal ID created by an agent will be reported back to you?
None of these require new technology on your side. They are the same disciplines that good representation has always relied on, and agents simply reward them more directly. If you want to see how this fits into a broader monetization approach, our piece on agentic monetization covers the AI-assisted decisions on the yield side.