Vercel

Vercel Connect and Enterprise Managed Users both hit GA, and eve agents can now be built directly from the Vercel dashboard. The dashboard builder scaffolds agent code, pushes it to your Git repo, and creates a project pre-wired with AI Gateway models, web chat or Slack, and MCP connections. Vercel Connect's secure connectivity layer for AI agents and apps is now generally available.

Vercel for Slack exited into public beta, bringing team coding, production investigations, and user-approved remediation actions into Slack.

Speed Insights now has a free tier, giving every plan a performance overview from real users. Paid capabilities continue as Speed Insights Plus. Deployment Storage is now billed at $0.10 per GB-month for new Pro and Enterprise teams, with retention controls in Deployment History.

AI Gateway expanded its model catalog. Gemini 3.8 Flash, Muse Spark 1.3, Claude Fable 5.1, Qwen 3.8 Max 0902, Tencent's Hy4 Preview, MiniMax H3 (50% off through September 13), and GLM-5.3 via DigitalOcean (50% off through September 7) all landed. Per-user budgets now cap individual spend alongside team, project, and API key budgets.

Platform surface expanded. Cursor Cloud Agents now run in Vercel Sandbox with scale-to-zero workers and isolated microVMs. Vercel CLI gained commands for DNS records, domains, projects, and members. Pro and Enterprise teams can select cheaper Basic build machines (2 vCPU, 8 GB). The free domain with Pro now includes .app and .dev. In the AI SDK, the fx adapter joined the harness layer.

Next.js 16.3.3 and 15.5.24 fixed two critical unauthenticated RCE vulnerabilities — one on Windows-hosted servers, one in the Image Optimization API via AVIF files. Turborepo 2.10.10–2.10.12 walks the repo once when pruning, supports nub.lock, and caches native uv tool tasks. In the AI SDK, Deepgram's diarize no longer defaults to true (paid add-on); Deepgram transcription options now pass through correctly.