Vercel AI Updates: November 14, 2026
Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled
1. Workflow SDK 5 Ships With Per-Run Encryption, Regional Runs and a Faster Engine for Agent Loops
Vercel released Workflow SDK 5 on November 12. Workflow and step inputs and outputs, hook payloads and stream data are now encrypted with AES-256-GCM under a unique key per run, with no setup on Vercel, and self-hosted Worlds can opt in by implementing getEncryptionKeyForRun(). Workflow infrastructure now runs in every Vercel Function region, each run has a home region that can be set with a region option on start(), and the trace viewer was redesigned with span search and searchable run attributes. Vercel cites agent loops and harnesses as a key use case and reports 1.6x to 25x improvements from an engine rewrite: in its own benchmark of v5.2.0 against v4.8.12 in iad1, p75 time between steps fell from 618 ms to 109 ms and a 1,020-step run went from a median of 21.0 minutes to 3.0. The release also brings a Python SDK in beta, WorkflowAgent in @ai-sdk/workflow, and a durable-harness package that supports Pi and OpenCode, with experimental Claude Code and Codex integrations in Vercel Sandbox. The post says v5 has no major breaking changes and that the benchmark covers one workload in one region. Source
2. Parallel Decisions Fast Joins AI Gateway, Free Through November 23
Vercel added Parallel Decisions Fast from Parallel to AI Gateway on November 13 under the model ID parallel/parallel-decisions-fast. The model searches the live web to answer typed questions and returns probabilities with cited evidence: boolean questions return the probability of “yes”, choice questions select from defined options, and score questions rate against an ordered rubric. It can be called through experimental_decide in the AI SDK or an OpenAI-compatible Decisions endpoint at ai-gateway.vercel.sh/v1/decisions, and no separate Parallel account is needed. The changelog suggests acting automatically above a probability threshold and routing uncertain results to human review, for tasks such as CRM enrichment, eligibility checks and claim verification. Usage is free through November 23, 2026; pricing after that date is not stated. Source
3. Token Pass Raises Its Monthly Allowance From $80 to $150 in Tokens at the Same $30 Price
Vercel increased the Token Pass allowance to $150 of tokens per month, up from $80, with the price unchanged at $30 per team per month. Token Pass is an Early Access add-on for Pro teams that covers featured open-weight models on AI Gateway: Kimi K3, Kimi K3.1 Pro, GLM 5.3, GLM 5.3 Flash and DeepSeek V4.1 Flash. The credit resets each billing period and does not roll over. When it runs out, requests made with standard AI Gateway keys or OIDC tokens fall back to the team’s AI Gateway balance, while requests made with Token Pass-scoped keys stop with no overage charges. Spots are limited and a waitlist applies once they fill. Source
4. Vercel Sandbox Starts Up to 4x Faster Outside iad1 Through Regional Image Caching
Vercel said Sandbox now initializes up to 4x faster outside the Washington, D.C. (iad1) region, because reused image data is read from a regional cache near where the sandbox runs and no longer requires round trips to origin image storage on cache hits. The caching applies to managed and custom images backed by Vercel Container Registry, including snapshots based on them. In Vercel’s internal end-to-end tests with Managed Images, startup in Seoul fell from about 3.5 seconds to 1.05 seconds and startup in Paris fell by about 75%. Regions closer to origin storage see smaller gains, and no action is needed to benefit. Source
5. Vercel Security Review Now Scans GitHub Repositories That Have No Linked Vercel Project
Vercel removed the requirement that Security Review, a feature enabled in Vercel Agent settings that scans code changes for vulnerabilities, run against a linked Vercel project. It now checks commits directly in repositories, and by default covers every unlinked repository the Vercel GitHub App can reach in a connected GitHub account, with each push triggering a review. Coverage can be narrowed to selected accounts or repositories in Agent settings. When a push also creates a deployment for a linked project with Security Review enabled, the review uses that project and deployment context as well. The changelog does not state pricing or plan availability. Source
6. Vercel Daemons Can Export OpenTelemetry Traces to Vercel Tracing
Vercel added trace export for Daemons, its background workers, so their operations can be inspected alongside the rest of a project’s traces by developers and their agents. Each daemon gets a Unix socket for OTLP/HTTP traces, exposed through the VERCEL_OTEL_SOCKET_PATH environment variable. Node.js daemons on the default configuration export automatically after upgrading @vercel/otel to 2.3.0 or later and calling registerOTel({ serviceName: 'my-daemon' }) before the worker starts. Spans can be queried from the CLI, for example vercel traces search --project my-project --service my-daemon --json. Source