Host Platform Architecture

This page summarizes the shared architectural responsibilities of any host platform that operates an ARTF execution environment. A "host platform" in this context is any system that runs the core auction or decisioning process and controls where and how ARTF Bidding Agents execute.

graph LR Media[Media and data owners] --> Host[ARTF Host Platform] Host --> Agent1[ARTF Bidding Agent] Host --> Agent2[ARTF Bidding Agent] Host --> Buyers[Buy side platforms]

1. Shared architecture and operations

Under ARTF, the host is the orchestrator of a containerized execution environment. It:

  • Runs the core auction or decisioning workflow.
  • Admits, schedules, and monitors agent containers.
  • Decides which data is provided to each agent and which mutations to accept.

This places the host at the center of:

  • Data governance. Managing how sensitive or high-value data is accessed and used.
  • Service orchestration. Coordinating multiple agents (identity, enrichment, fraud, measurement, bidding) within strict latency budgets.

2. Performance and scaling benefits

Legacy architectures rely on multiple external HTTP calls for enrichment, identity, and fraud checks. Each call adds latency and introduces variability. ARTF addresses this by:

  • Running agent services as co-located containers inside the host's infrastructure.
  • Using high-performance remote procedure calls (such as gRPC) for host–agent interaction.

For hosts, this can translate into:

  • Lower and more predictable end-to-end latencies.
  • The ability to run more complex logic within the same auction timeout.
  • Improved handling of high-demand scenarios, including live events and other latency-sensitive supply.

3. Standardized integration surface for partners

Historically, each new integration (for identity, enrichment, fraud, or curation) has required bespoke engineering work. ARTF introduces:

  • A standard container model for agents, including manifests that describe capabilities and requirements.
  • A standard interaction model based on OpenRTB Patch and intents.

This allows hosts to:

  • Offer a consistent, documented environment for partners to deploy their agents.
  • Reduce the number of custom point integrations to maintain.
  • Treat their infrastructure as a reusable, extensible platform.

4. Enabling secure collaboration with data owners

For hosts working with publishers, broadcasters, or retail media networks, data protection is often the primary constraint. ARTF supports:

  • Running agents inside the same secure environment where sensitive data resides.
  • Sharing only the fields or derived signals that are necessary for a given intent.

This can make it more practical to:

  • Activate first-party segments or contextual data in real time without expanding the external data surface.
  • Collaborate with sensitive-data verticals that were previously reluctant to participate in open auctions.

5. Operational responsibilities

Adopting ARTF changes the operational profile of host platforms. In addition to existing responsibilities, hosts must:

  • Manage container infrastructure. Provision resources, enforce quotas, and maintain reliability for many third-party agents.
  • Enforce policies. Determine which agents are admitted, which intents they may declare, and how conflicts between mutations are resolved.
  • Monitor behavior. Track which agents ran on which impressions, what mutations they proposed, and which changes were applied.

These responsibilities are central to ensuring that ARTF delivers its intended performance and governance benefits.

6. Migration and readiness

Host platforms considering ARTF adoption should plan for:

  • Architecture review. Assessing how current enrichment, identity, and fraud workflows can be migrated from external calls to containerized agents.
  • Capacity planning. Estimating the resource impact of running agents in-house versus calling out to external services.
  • Partner engagement. Communicating with key vendors and buy-side partners about ARTF support, timelines, and requirements.

Platforms that have already invested in container orchestration, service meshes, and observability will typically find it easier to take advantage of ARTF, and may realize performance and data-collaboration advantages earlier than others.

7. Positioning and incentives

ARTF is defined as an execution model that complements, rather than replaces, existing host businesses:

  • Preserves commercial roles. Hosts continue to own the relationship with publishers, buyers, and partners; ARTF standardizes how technical integrations are executed, not how commercial value is shared.
  • Enables differentiation. A host can curate which agents to support, how they are orchestrated, and what additional services are offered around them, maintaining room for product and strategy differentiation.
  • Aligns incentives. Lower latency, better data governance, and more reliable integrations are beneficial to hosts, publishers, buyers, and technology partners, reducing friction rather than creating new points of conflict.