Skip to main content
Software Engineering Daily

Scaling Agent Workloads at Vercel

51 min episode · 2 min read
·
Andrew Barba,Shar Dara

Episode

51 min

Read time

2 min

Topics

Startups, Marketing, Artificial Intelligence

AI-Generated Summary

Key Takeaways

  • Declarative Agent Architecture: Eve agents are defined through a directory of markdown and YAML files — starting with a single `instructions.md` — that compile into provisioned infrastructure. Adding a `skills/` folder automatically provisions a sandbox; adding `schedules/` triggers event infrastructure. This means zero manual infrastructure configuration for standard agent deployments.
  • Progressive Skill Disclosure: Rather than loading all context upfront, Eve uses a `skills/` directory where each skill includes a hint describing when it should activate. The model pulls additional context only when needed, preventing context bloat. Vercel's dZERO agent is approximately 90% skills-based, keeping base context lean while maintaining broad capability coverage.
  • LLM-Friendly Documentation as a Development Strategy: Eve ships LLM-readable docs and a changelog inside the package itself. Coding agents like Codex detect version differences — for example, jumping from v0.2 to v0.5 — and read only the changelog diff rather than re-learning the full framework. This approach is critical for pre-1.0 frameworks that break APIs weekly.
  • Eval-First Testing for Nondeterministic Systems: Eve provides two eval types: deterministic condition checks (verifying specific tools were called) and LLM-as-judge scoring for qualitative outputs like SQL query correctness. Eve's own test suite runs entirely as evals, enabling multiple daily releases. Developers can trigger human-in-the-loop steps and full agent turns programmatically within eval functions.
  • Scoped Memory Architecture with Provider Adapters: Eve's forthcoming memory system defines named slots with explicit scopes — global workspace, per-user, or per-channel-plus-user — preventing agents from retrieving memories belonging to other users. A provider contract supports implementations from simple file storage to services like SuperMemory, which performs agentic recall based on inbound message content rather than static file retrieval.

What It Covers

Andrew Barba and Shar Dara from Vercel explain Eve, an open-source cloud-native agent framework built from lessons scaling their internal dZERO data agent. Eve compiles declarative markdown and YAML configuration files into infrastructure-as-code, enabling horizontally scalable, session-isolated agents serving thousands of concurrent users across enterprise and consumer contexts.

Key Questions Answered

  • Declarative Agent Architecture: Eve agents are defined through a directory of markdown and YAML files — starting with a single `instructions.md` — that compile into provisioned infrastructure. Adding a `skills/` folder automatically provisions a sandbox; adding `schedules/` triggers event infrastructure. This means zero manual infrastructure configuration for standard agent deployments.
  • Progressive Skill Disclosure: Rather than loading all context upfront, Eve uses a `skills/` directory where each skill includes a hint describing when it should activate. The model pulls additional context only when needed, preventing context bloat. Vercel's dZERO agent is approximately 90% skills-based, keeping base context lean while maintaining broad capability coverage.
  • LLM-Friendly Documentation as a Development Strategy: Eve ships LLM-readable docs and a changelog inside the package itself. Coding agents like Codex detect version differences — for example, jumping from v0.2 to v0.5 — and read only the changelog diff rather than re-learning the full framework. This approach is critical for pre-1.0 frameworks that break APIs weekly.
  • Eval-First Testing for Nondeterministic Systems: Eve provides two eval types: deterministic condition checks (verifying specific tools were called) and LLM-as-judge scoring for qualitative outputs like SQL query correctness. Eve's own test suite runs entirely as evals, enabling multiple daily releases. Developers can trigger human-in-the-loop steps and full agent turns programmatically within eval functions.
  • Scoped Memory Architecture with Provider Adapters: Eve's forthcoming memory system defines named slots with explicit scopes — global workspace, per-user, or per-channel-plus-user — preventing agents from retrieving memories belonging to other users. A provider contract supports implementations from simple file storage to services like SuperMemory, which performs agentic recall based on inbound message content rather than static file retrieval.

Notable Moment

When asked about self-evolution, Barba described agents opening pull requests against their own codebase via a coding sub-agent. This model lets teams configure any level of human oversight — from auto-merge to mandatory review — using standard Git workflows rather than custom agent governance tooling.

Know someone who'd find this useful?

Episode Transcript

Most AI agent setups today are built around a single session, where one user interacts with one agent at a time. However, that model breaks down when an agent has to serve a business, where thousands of requests can arrive at once and each session needs to be isolated, durable, and recoverable. Getting agents to run reliably at that scale has meant a lot of hand rolled infrastructure beneath the agent itself. Eve is an open source, cloud native agent framework from Vercel that removes much of the agent scaling burden. In the Eve framework, an agent is defined declaratively through configuration files, and these files compile into infrastructure as code, so the platform provisions only what the agent actually uses. Andrew Barba is a member of technical staff at Vercel, and Shar Dara is the product lead for Eve at Vercel. In this episode, they join Kevin Ball to discuss what it means for an agent framework to be cloud native, why they chose to express agents in plain English, and their view that company building is becoming agent building. Kevin Ball, or Kate Ball, is the vice president of engineering at Mento and an independent coach for engineers and engineering leaders. He co founded and served as CTO for two companies, founded the San Diego JavaScript Meetup, and organizes the AI in Action discussion group through Latent Space. Check out the show notes to follow k ball on Twitter or LinkedIn, or visit his website, kball.llc. Andrew, Shar, welcome to the show. Thanks so much for having us. Happy to be here. Yeah. Thanks for having us. I'm excited to get to know you guys and to learn a little bit more about Eve and this new AgenTic framework, but let's start with you. So we'll go to Andrew first. Can you tell us a little bit more about your background and how you ended up working on this project? Yeah. So I've been at Vercel for some three and a half years now. I was actually a Vercel customer, maybe one of Vercel's very first enterprise customers back in 2020. So bit of an unconventional path to Vercel. But, yeah, I spent my first year and a half or so doing all CDN works, a lot of, like, deep systems, distributed systems, edge networks, firewalls, security stuff, things like that. And then basically went over to the other end of the business and worked on billing for a good year, which is where I met Shar. So Shar and I were leading we're leading the billing team at Vercel for most of 2025. And then, yeah, February came around this year, and we had a really interesting opportunity to kinda look at what does it take to deploy agents to Vercel, and we came up with Eve. Awesome. And Shar, how about you, your background, and how you ended up here? Yeah. I've been at Vercel for a year and a half. I was in the billing …

Get the full transcript (10,281 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all Software Engineering Daily transcripts →

You just read a 3-minute summary of a 48-minute episode.

Get Software Engineering Daily summarized like this every Monday — plus up to 2 more podcasts, free.

Pick Your Podcasts — Free

Keep Reading

Books, tools, and gear mentioned in this episode

SignalCast may earn commission on purchases via these links.

Tools

  • by Vercel

    Andrew Barba and Shar Dara from Vercel explain Eve, an open-source cloud-native agent framework built from lessons scaling their internal dZERO data agent.
  • A provider contract supports implementations from simple file storage to services like SuperMemory, which performs agentic recall based on inbound message content rather than static file retrieval.
  • SPONSORS: Warp Build
  • SPONSORS: GuardSquare
  • SPONSORS: Fingerprint

More from Software Engineering Daily

We summarize every new episode. Want them in your inbox?

Similar Episodes

Related episodes from other podcasts

Explore Related Topics

This podcast is featured in Best Cybersecurity Podcasts (2026) — ranked and reviewed with AI summaries.

Read this week's Startups & Product Podcast Insights — cross-podcast analysis updated weekly.

You're clearly into Software Engineering Daily.

Every Monday, we deliver AI summaries of the latest episodes from Software Engineering Daily and 192+ other podcasts. Free for one show.

Start My Monday Digest

No credit card · Unsubscribe anytime