Changelog

The AI SDK Harness adds two new adapters, Deep Agents and OpenCode, so you can run either coding-agent runtime through one unified interface inside a Vercel Sandbox.

New projects on Vercel now authenticate Flags automatically — no SDK Keys or environment variables to configure.

Pro now supports up to 500 concurrent builds on Vercel, up from 12. Large repos and monorepos ship faster without waiting on queued builds.

GLM 5.2 Fast via Wafer AI is now available on Vercel AI Gateway. Use GLM 5.2 via Wafer's fast inference stack for throughput averaging 170+ tokens per second.

Node APIs can now be deployed on Vercel with zero-configuration, with fluid compute and automatic scaling

Diagnose domain DNS configuration directly from the Vercel CLI with vercel domains verify. Check DNS status, project attachment, and get exact fix instructions when something is wrong.

Start a Chat SDK bot with one command. create-chat-sdk scaffolds a Next.js project with your platform and state adapters, env vars, and a webhook route.

The Vercel CLI now preserves .env.local when running vercel link, only updating VERCEL_OIDC_TOKEN without overwriting your other local environment variables.