Vercel ships agent controls while ServiceNow ships a guardrail for them

AI SDK 6, MiniMax M2.1 on AI Gateway and GLM-4.7 on AI Gateway, and The Deploy Log for this week. Every price and claim read off the publisher's own page, with a call on each.

Compiled in September 2026 from the week's sources and dated to the week it covers. It was not sent by email.

The week in The Deploy Log, every row sourced

Vercel

AI SDK 6

DEPLOYif you build TypeScript agents

WHAT SHIPPEDAI SDK 6 adds an Agent abstraction, ToolLoopAgent with a 20-step default loop, tool execution approval via a needsApproval flag, strict mode, MCP tool calling with structured output, DevTools, reranking, standard JSON Schema, and image editing. Thomson Reuters built CoCounsel with 3 developers in 2 months and now serves 1,300 accounting firms.

THE CATCHThe approval flag is the safety layer: set needsApproval true for destructive tools like rm -rf, or pass a function to decide per input. The same tool definition powers agent logic, API responses, and typed UI components.

IF YOU RUNa Next.js or React app with tool calling agents and want human approval before file deletion or payment processing

SOURCEVercel, Introducing agents, tool execution approval, DevTools, full MCP support, reranking, image editing, and more


Vercel

MiniMax M2.1 on AI Gateway

DEPLOYif you route coding models through Vercel

WHAT SHIPPEDMiniMax M2.1 is live on Vercel AI Gateway with no other provider accounts required. It is faster than M2 with improvements in coding, multi-step tool call tasks, refactoring, feature adds, bug fixes, and code review across Go, C++, JS, C#, and TS.

THE CATCHSet the model to minimax/minimax-m2.1 in the AI SDK. The Gateway adds unified API, usage and cost tracking, retries, failover, observability, and Bring Your Own Key support.

IF YOU RUNa multi-provider AI stack and want one endpoint for MiniMax M2.1 with failover and cost tracking

SOURCEVercel, MiniMax M2.1 now live on Vercel AI Gateway


Vercel

GLM-4.7 on AI Gateway

DEPLOYif you build agentic front ends

WHAT SHIPPEDZ.ai GLM-4.7 is live on Vercel AI Gateway with no other provider accounts required. It brings major improvements in coding, tool usage, multi-step reasoning, complex agentic tasks, conversational tone, and front-end aesthetics.

THE CATCHSet the model to zai/glm-4.7 in the AI SDK. The Gateway provides unified API, usage and cost tracking, retries, failover, observability, and Bring Your Own Key support.

IF YOU RUNa Vercel app that needs a single endpoint for GLM-4.7 with automatic retries and provider routing

SOURCEVercel, GLM-4.7 available on Vercel AI Gateway

The Deploy Log

Models and APIs

DEPLOYif you run production LLM agents
AprielGuard. ServiceNow released AprielGuard, an 8B parameter safety and security safeguard model that detects 16 safety risk categories and adversarial attacks including prompt injection, jailbreaks, chain-of-thought corruption, context hijacking, memory poisoning, and multi-agent exploit sequences across standalone prompts, multi-turn conversations, and agentic workflows. Hugging Face

Tools and products

WATCHif you build branded image generation demos
Pixel Portraits template. Vercel released a v0 template that turns a selfie into a 16-bit pixel trading card using GPT Image 1.5 Edit, BRIA RMBG 2.0 background removal, Canvas API compositing, and a custom LoRA trained on pixel art, with the on-site pipeline completing in roughly six seconds. Vercel

Left out

  • Function start type in Runtime Logs: its own name for the thing was too long to print

The pattern

The week splits into two halves of the same production problem. Vercel gives builders the controls to make agents act: AI SDK 6 adds approval gates, typed tool loops, and two new gateway models for coding and agentic tasks. ServiceNow answers the question those controls raise: once an agent can delete files and call tools, what watches the agent itself?

AprielGuard is an 8B model trained on synthetic agentic workflows with corrupted reasoning traces, tool outputs, and memory states, and it runs in reasoning or fast mode. The deployment lens says the stack is now complete enough to ship: Vercel for the action layer, ServiceNow for the guardrail layer, and a real decision about which tools get the needsApproval flag before the first rm -rf.

Follow the wire

One line every time AI actually ships, around the clock, free on Threads and Bluesky. Every one is in The Deploy Log with its source.

Follow on Threads Follow on Bluesky

The Saturday edition is the week in one, free to read here. By email if you want it, every Saturday at 7:00 AM ET:

Every edition