Designing products for AI agents: Resend's approach to agent UX and discovery

Latent Space interviews Resend co-founder Zeno Rocha on designing products for AI agents. The discussion centers on answer engine optimization (AEO)—how agents like Claude Code discover and choose APIs—using Resend's email service as a case study. Rocha walks through concrete examples: how agents evaluate providers, what makes an API agent-discoverable, and the product decisions required to let agents register accounts, read pricing, and execute actions without human intervention.

The conversation covers technical challenges including email deliverability, DNS configuration, tracking pixel accuracy, and abuse prevention. Rocha explains why serving agents requires different product thinking than serving humans: agents need structured, machine-readable information; pricing in Markdown; and 80+ named MCP (Model Context Protocol) tools for fine-grained capability exposure.

Key themes include building an agent-first email stack, standardizing agent registration, and measuring adoption across model versions. The discussion treats agent product design as a distinct discipline emerging from the convergence of AI models, API design, and user experience.