The DropJune 25, 2026via Vercel Blog
AI SDK 7
Why it matters
AI SDK 7 moves agent development from experimental to production-grade by adding durability (resumable workflows surviving restarts), fine-grained tool approvals with human-in-the-loop, and enterprise observability. This is critical infrastructure for teams scaling beyond chatbots to long-running, autonomous agent systems.
Key signals
- 16 million weekly downloads for AI SDK
- AI SDK 7 adds WorkflowAgent for durable, resumable execution across process restarts and interruptions
- Tool approval system supports automatic, human-in-the-loop, and HMAC-signed approvals
- First-class timeout configuration: total, per-step, per-chunk, per-tool limits
- Telemetry integrations: Datadog, Langfuse, Braintrust, Raindrop, Sentry, Laminar, Langsmith
- HarnessAgent enables unified API to run Claude Code, Codex, Pi, Deep Agents, OpenCode
- Experimental provider-agnostic realtime support (OpenAI, Google, xAI)
- Video generation support (fal, Google AI Studio, Vertex, Replicate)
- MCP Apps support with model-visible vs app-only tool separation
- SandboxSession abstraction for portable command execution across dev/CI/production
- Performance statistics per step: token arrival speed, model response latency, tool execution time
The hook
16M weekly downloads. Vercel just shipped AI SDK 7 with agent durability, tool approvals, and observability—production-ready infrastructure for autonomous workflows.
AI SDK, with over 16 million weekly downloads, is the TypeScript SDK for building AI applications, features, frameworks, and agents across any model provider. It's the same layer , Vercel's open-source agent framework, is built on.eve
AI SDK 7 adds production depth for agent work across five areas:
Building well-behaved agents requires fine-grained control over model reasoning, tool context, and file handling.
Most frontier models support configurable reasoning, but every provider API exposes it differently.
AI SDK 7 standardizes this with a option for and . It maps to provider-native reasoning settings, letting you control reasoning effort in a single line. You can also still fall back to provider options when you need more detailed provider-specific reasoning configuration.reasoninggenerateTextstreamText
Learn more in the .reasoning documentation
Tools are increasingly developed independently of specific agents or applications. For example, third-party companies offer tools that enable agents to use their APIs. Therefore, tools require additional inputs that are not generated by LLMs, such as API keys or configuration settings.
AI SDK 7 adds a fully typed tool context that can be specified for each tool via a schema. The context is limited to the tool to prevent 3rd-party tools from accessing context they do not need.
Learn more about Tool Context
For more complex agentic loops, you often need variables that you can access and modify in to adjust prompts, model selection, and more.prepareStep
AI SDK 7 introduces a typed runtime context available during step preparation and tool approval functions, with optional telemetry support. This enables you to encapsulate more logic in and share those agents with that internal logic.ToolLoopAgent
Learn more about .Runtime Context
Many agent workflows require handling large inputs, such as PDFs, images, datasets, or other artifacts. Sending those files inline is slow and wasteful, especially for stateless inference, where...