Sign InOpen Brain
VercelEngineering PostOfficial Source

How Featured's users make 100K media pitches per month on Vercel

A three-engineer team consolidated model routing, long-running agent jobs, and deployment on Vercel while serving 100,000+ users. The case study shows where managed abstractions can remove operational work.

Vercel · Sep 11, 2026
Open Source Open MarkdownOpen JSON
Source Summary

Featured supports **100,000+ users** across three brands with three engineers. It migrated **374 Sanity sites** from Elastic Beanstalk, routes model traffic through AI SDK and AI Gateway, and moved long-running jobs to Workflow SDK.

Practical Implication

For a small agent team, the useful pattern is consolidation: standardize model calls, make provider changes configuration-level, and put work that outlives a request into durable workflows. Featured also made chat primary after users favored conversation over dashboard navigation.

Agent-Ready Context
Featured supports **100,000+ users** across three brands with three engineers. It migrated **374 Sanity sites** from Elastic Beanstalk, routes model traffic through AI SDK and AI Gateway, and moved long-running jobs to Workflow SDK.

For a small agent team, the useful pattern is consolidation: standardize model calls, make provider changes configuration-level, and put work that outlives a request into durable workflows. Featured also made chat primary after users favored conversation over dashboard navigation.

This is a vendor-authored case study, not a cost or reliability comparison. Its **17-model** setup and **100,000+ monthly pitches** show operating scale, but the material provides no latency, failure-rate, or migration-cost data.
Connected Context · Feed7 Judgment

This turns gateway abstraction and durable workflows from product features into an operating pattern used by a three-engineer team across hundreds of sites and many models. It reinforces consolidation as a way to limit integration burden at scale, while narrowing the evidence to feasibility: without cost, latency, failure, or migration data, it does not establish that the architecture is efficient or superior.

How Searchable ships customer-requested features in 30 minutes on VercelSearchable independently reinforces the implementation benefit of keeping model swaps behind AI SDK and Gateway, though its evidence concerns delivery speed rather than Featured’s operating scale.Introducing the Agents APIThe managed Agents API addresses the same need for persistent, orchestrated work that Featured moved into durable workflows, but places more of the runtime inside a hosted agent service.The Agent Behind the Curtain: Building the Oz Cloud Agent Platform — Safia Abdalla, WarpWarp’s separation of runtime, harness, artifacts, and orchestration provides a broader portability model for the consolidation pattern Featured applies through shared model routing and workflows.How Tailscale built a customer-facing model router on AI GatewayTailscale adds the identity, retention, and execution-isolation controls that become implementation requirements when Featured’s shared gateway pattern is used for customer-facing agents.
Context Map
infracoding#gateways#cloud-agents#agent-sdks
Uncertainty
This is a vendor-authored case study, not a cost or reliability comparison. Its **17-model** setup and **100,000+ monthly pitches** show operating scale, but the material provides no latency, failure-rate, or migration-cost data.