Wednesday, November 4, 2026 Edition No. 228
Search
Daily News · 4 min read

Vercel AI Updates: November 3, 2026

1. Vercel MCP Accepts Any MCP Client That Publishes a Client ID Metadata Document

Vercel opened its hosted Vercel MCP server to any MCP client that identifies itself with a Client ID Metadata Document (CIMD), a JSON file on the client’s own domain whose HTTPS URL is used as the OAuth client_id. Previously Vercel reviewed and allowlisted each client and its redirect URIs by hand and rejected anything not on the list; users still log in and approve every connection, and apps already registered with Vercel are unchanged. The change follows the MCP authorization spec, which now recommends CIMD over Dynamic Client Registration. Because CIMD clients are not reviewed and do not get the Vercel verified badge, the consent screen shows each client’s name, logo and the hostname serving its metadata document (for example “Claude Code by claude.ai”), so that hostname is what a user is trusting when approving. Source

2. Vercel MCP Renders Deployments as an Interactive Card in ChatGPT and Other MCP Apps Clients

Vercel shipped its first MCP App: Vercel MCP can now show deployments as an interactive card inside ChatGPT and other chat clients that support the MCP Apps extension. The card shows live deployment status, build logs, and source and commit details, adds a Visit button once the deployment URL is ready, lets users send logs into the chat to debug with the model or open a fullscreen log view, and offers a project list for repositories linked to multiple projects. The underlying open_deployments tool looks up deployments by project, repository or commit with environment filtering, and clients without MCP Apps support receive the deployment state, preview URL and commit as text instead. ChatGPT connects to Vercel MCP through its Connectors setup. Source

3. Meituan’s LongCat 2.5 Arrives on AI Gateway With a 1M-Token Context Window

Vercel added Meituan’s LongCat 2.5 to AI Gateway under the model name meituan/longcat-2.5. The model takes text and image inputs with a 1M-token context window and combines native visual understanding with coding and multi-step tool use; Meituan highlights gains in visual programming, creative design and 3D tasks, along with longer agent workflows that carry engineering work from planning through execution. It can be selected in Claude Code, Codex, Cursor and other agents after installing the latest Vercel CLI and running its AI Gateway setup, or tried in the model playground, and AI Gateway passes through provider pricing with no markup and no platform fee on inference, including BYOK requests. Source

4. Vercel Sandboxes With Drives Now Start in 0.7 Seconds at the Median

Vercel cut creation time for Vercel Sandboxes that mount Drives, its persistent storage for sandboxed agents, from a median of 1.7 seconds to 0.7 seconds, with p95 dropping from 2.2 to 1.6 seconds. The speedup applies automatically to every sandbox that uses Drives, with no SDK upgrade or configuration change, and pricing is unchanged. For agent workloads that create a fresh sandbox per task while keeping state on a Drive, this shrinks the startup cost of that pattern. Source