Harbor on Vercel Sandbox: parallel Terminal-Bench without your laptop

UpdatedSeptember 18, 2026

Harbor is the open-source eval harness behind Terminal-Bench. As of 17 September 2026, harbor run --env vercel places each trial in its own Firecracker microVM on Vercel Sandbox. That is the production change: you can parallelize agent benchmarks beyond your laptop without collapsing network policy and secrets into the same guest as the agent under test.

If you already measure coding agents with a fixed bench, this update changes where the trials run — not whether you still need a written acceptance gate. Pair it with a durable evaluation loop, not a one-off screenshot of pass rates.

This is an excerpt. Read the full post at otf-kit.dev/blog/harbor-evals-vercel-sandbox — full-stack kits your AI coding agent can actually ship to production. Browse the kits →

#agents #vercel #ai-tools

Comments

Join the discussion

No comments yet. Be the first to comment.

More from this blog

Marketplace kit: buyer hire flow, seller workspace, and Stripe checkout you own

A two-sided services marketplace is not an appointments calendar and it is not a fitness member app. After you buy the Marketplace kit, you get a buyer path (browse, hire, Stripe checkout, orders, ord

Sep 18, 20261 min read

Marketplace kit: buyer hire flow, seller workspace, and Stripe checkout you own

Better Auth on Expo: SecureStore sessions, cookiePrefix match, getCookie APIs

Better Auth’s Expo client already solves the hard part of native sessions: it caches session data in expo-secure-store, filters cookies with cookiePrefix, and exposes getCookie() so your own API calls

Sep 18, 20261 min read

Better Auth on Expo: SecureStore sessions, cookiePrefix match, getCookie APIs

Everything Bundle vs one $99 kit: when the $149 pack pays for itself

You already know you need a paid kit. The free SDK covers components. Checkout is open. The real hesitation is narrower: buy one $99 lane, or pay $149 once for the Everything Bundle. Buy the Everyth

Sep 18, 20261 min read

Everything Bundle vs one $99 kit: when the $149 pack pays for itself

Free SDK vs paid kit: when buying actually pays for itself at checkout

You already run the free SDK. Screens look decent. Components render the same on web and native. Then you open checkout for a full-stack kit and freeze — not because the product is unclear, but becaus

Sep 18, 20261 min read

Free SDK vs paid kit: when buying actually pays for itself at checkout

Keep unsigned users off product screens with one Expo Router root gate

If a product screen can mount before you know who is signed in, the file tree is doing the wrong job. The fix is not a Redirect at the top of every screen. It is a file convention that splits signed-o

Sep 18, 20261 min read

Keep unsigned users off product screens with one Expo Router root gate

OTF — kits your AI coding agent can ship to production

609 posts

Engineering notes on shipping production apps with AI coding tools — Claude Code, Cursor, Codex, Lovable, Bolt — and the stack underneath: React Native, Expo, Next.js, Supabase. Honest takes on what works, what breaks, and the full-stack kits that get you to production faster. By OTF.

Command Palette

Comments

More from this blog