AI SDK 7 is now available
AI SDK 7 just shipped. Agents now run durable, observable, and production-ready—without the demo-to-deploy chasm.
Why it matters
Vercel's AI SDK 7 moves agent development from experimentation to production infrastructure. The release adds durable execution, reasoning control, tool approvals, sandbox support, and observability—the primitives teams need to ship agents that survive restarts, scale across text/audio/video, and integrate with existing coding agents (Claude Code, Codex, Pi).
The key facts
10 to knowDurable execution with state persistence across deploys and interruptions
Provider-agnostic reasoning control across OpenAI, Anthropic, Google, Groq, xAI, Bedrock, Fireworks, DeepSeek
Tool approval policies at call and agent levels with HMAC signing for tamper-proof resumption
Harness layer to wrap Claude Code, Codex, Pi behind unified agent interface
Realtime voice (experimental) and video generation (experimental) support
OpenTelemetry package (@ai-sdk/otel) with Node.js tracing channel integration
Terminal UI (@ai-sdk/tui) for interactive agent testing before production
Node.js 22+ requirement; ESM imports only (CommonJS deprecated)
MCP protocol version 2025-11-25 with app-specific rendering and sandboxed iframe support
First-class timeout budgets (total, per-step, per-chunk, per-tool)
Go to the source
Vercel Blogvercel.com
Publisher excerpt: AI SDK 7 is a major release for building production agents in TypeScript. The SDK has grown from model calls and chat primitives into a broader agent platform for developing, running, integrating, and observing agents across text, audio, realtime, image, and video. Every major provider is supported…