Vercel — Cloud, Developer Tools product updates and releases, tracked on megachangelog.
You can now create and deploy agents directly from the Vercel dashboard with a builder that scaffolds the project, creates a private Git repository, and deploys it to your team. Configure the agent's identity, choose an AI model through AI Gateway, add a web chat or Slack interface, and connect tools and data from services like Linear, Notion, or custom MCP servers.
Vercel CLI now includes dedicated commands for managing DNS records, domains, and projects, enabling terminal-based configuration that was previously only available in the dashboard. New capabilities include updating DNS records, renewing domains, controlling project state and observability, and managing project members with JSON output support.
Hy4 Preview, a 770B parameter open-source Mixture-of-Experts model from Tencent, is now available on Vercel's AI Gateway. The model supports 1M token context windows and is optimized for coding, document analysis, and scientific reasoning, accessible via the AI SDK with no platform fees.
The AI SDK harness layer now supports Cursor through an official adapter, allowing applications to switch between different coding agents without changing application code. The adapter uses the Agent Client Protocol to connect Cursor to the HarnessAgent interface.
The Deployments page now features redesigned filters that enable faster discovery of specific deployments. Users can apply suggested filters for common searches, type to find options, and use natural language to describe what they're looking for for automatic filter application.
Vercel introduces Workflow SDK, a new durable execution framework that lets developers write workflows as normal sequential code without complex DAG configuration or infrastructure management. The open-source, spec-based system uses only standard infrastructure like databases and queues, with client-side execution and automatic versioning to handle code changes without breaking in-flight runs.
Inclusion AI's Ling 3.0 Flash Fin, a finance-focused LLM with 256K token context and 32K output tokens, is now available on Vercel's AI Gateway at no cost through September 25. The model supports reasoning and function calling for financial research and multi-step analysis tasks.
GLM 5.3 Flash from Zhipu AI, a multimodal model with 1M token context window supporting text and vision input, is now available on Vercel AI Gateway. The model supports function calling, structured output, and streaming through the unified API interface.
Meta's Muse Image, their first image generation model, is now available on Vercel's AI Gateway. The model supports both image generation from text prompts and image editing with instructions, unified in a single model without switching.
Google's Gemini 3.5 Transcribe model is now available on AI Gateway, supporting both streaming and batch audio transcription with multi-language detection and custom vocabulary. The AI SDK V7 adds streamTranscribeReadable for real-time streaming transcription via WebSocket.
Alibaba's Qwen 3.8 Flash model is now available on Vercel's AI Gateway, supporting text and image inputs with a 1M token context window and up to 65k token responses. The model is recommended for coding, tool use, and multi-step agent workflows.
Python projects can now use routing rules to set response headers or rewrite requests to internal paths. Rules are evaluated by the Vercel CDN before requests reach your application and take effect immediately without requiring a new deployment.
The Vercel Security Dashboard is now generally available on all plans, automatically flagging security misconfigurations across accounts and projects with risk-ordered findings and actionable fixes. The dashboard is accessible via the Vercel UI and CLI, with support for muting noise, exporting results to CSV, and agentic remediation through the vercel security check command.
Two critical vulnerabilities affecting Next.js image optimization and Windows-hosted servers were disclosed in August 2026. Vercel-hosted applications are automatically protected and require no action, while self-hosted Next.js applications must upgrade to patched versions.
Speed Insights now offers a free tier providing high-level performance monitoring from real users across all plans with 10,000 events per team per month. The paid tier is renamed Speed Insights Plus with deeper diagnostics and historical data, while existing customers see only name changes with no impact to capabilities or pricing.
Vercel Connect replaces long-lived credential tokens with short-lived, scoped tokens requested at runtime, eliminating the security risks of stored secrets. The platform provides 100+ pre-built connectors for services like Slack, GitHub, and Salesforce, along with fine-grained RBAC, audit logging, and trigger-based event handling without exposing secrets to applications.
Vercel introduced the Run SDK, a package for safely executing untrusted JavaScript and TypeScript code in a sandboxed environment without direct access to application secrets or systems. It enables agents to write programs that call narrowly scoped host functions, with support for interrupting execution for authentication and human approval before resuming without repeating completed work.
Vercel's Chat SDK Slack adapter now supports Slack Enterprise Grid, enabling org-wide bot installations across workspaces with correct token resolution, tenant-scoped caches, and event deduplication. The adapter stores enterprise-wide installations by enterprise ID and handles multi-workspace deployments securely.
AI Gateway can now run video generation asynchronously using webhooks, polling, or deferred job retrieval, allowing applications to avoid blocking HTTP requests that could exceed timeouts. This enables more flexible workflows where your application can receive completion notifications or check status later instead of waiting for potentially long-running video renders.
MiniMax M3 and M2.7 models are now available for free on AI Gateway via GMI Cloud through September 6. Use the -free model IDs or set GMI Cloud as your preferred provider to access these models across all AI Gateway APIs.
Vercel's Chat SDK now includes a Notion adapter, allowing agents to participate in Notion page comments and discussions using the same codebase as other platforms like Slack and Discord. The adapter supports mentions, message editing, conversation history, and file attachments.
Alibaba's Wan 3.0 video generation model is now available on Vercel's AI Gateway. The unified model supports text-to-video, image-to-video, and reference-based generation with improved capabilities including up to 30 seconds at 1080p, and introduces three new asynchronous generation modes (poll, webhook, and fire-and-forget) to better support serverless and batch workflows.
The Chat SDK now includes an XChat adapter enabling bots to build end-to-end encrypted 1:1 and group conversations. The adapter automatically handles encryption, key management, and signature verification, with fallback rendering for XChat's limited markdown support.
Vercel Connect introduces a managed Linq connector enabling apps and agents to send and receive messages over iMessage, RCS, and SMS. Users can create a new Linq account or link an existing one without managing credentials directly.
Vercel Connect is now generally available on all plans, enabling secure runtime authentication with short-lived, scoped tokens instead of storing long-lived provider secrets. It includes 100+ preset connectors, per-environment attachment, audit logs, and supports AI SDK, MCP servers, and webhook triggers without stored secrets.
Elastic build machines now consider Turborepo cache hits when deciding machine size, preventing downgrades during warm-cache builds that could fail subsequent cold-cache builds due to insufficient resources. This change applies automatically to all builds with no action required.
Vercel Sandbox now runs in four regions globally (Washington D.C., San Francisco, Cleveland, and Paris), with support for all Vercel regions coming soon. Users can select regions close to their databases and services to reduce latency, and Pro and Enterprise teams can configure failover regions for automatic failover when the primary region is unavailable.
Vercel Functions on Bun runtime now support larger package sizes up to 5GB uncompressed and extended max duration up to 30 minutes (1800 seconds) for Pro and Enterprise teams. These features previously were only available on Node.js and Python runtimes.
Environment variables now use Config and Secret types instead of the Sensitive toggle, with automatic migration of existing Sensitive variables. A new Separate Production Secret Values policy is available, and the legacy Enforce Sensitive Environment Variables policy is deprecated.
Deployment Storage is a new feature that keeps deployment files available for inspection and instant rollback to previous versions within seconds. The feature is billed at $0.10 per GB per month with Hobby teams receiving 10 GB included, and users can customize retention policies for different deployment types to control storage costs.
Ora, an AI agent benchmarking platform, now runs on Vercel and benchmarks major AI agents including Claude Code, ChatGPT, Gemini, and Vercel's own eve framework. The platform shows customers where and why AI agents fail on live websites and measures performance metrics like steps to completion, success rates, and endpoint validity.
OpenAI lowered list pricing for GPT-5.6 Sol, and Vercel's 50% AI Gateway discount now applies to the new lower price through September 18. Input token costs drop 20% and output costs drop 33% across all service tiers, with automatic billing at new rates for existing requests.
The Vercel CLI now includes dedicated commands for managing DNS records, domains, and projects, bringing dashboard and API functionality to the terminal. Users can update DNS records by ID, renew domains, pause/resume projects, configure observability features, and manage project members—all with JSON output support for scripts and agents.
DeepSeek V4 Flash Vision Experimental is now available on Vercel's AI Gateway, enabling users to submit images alongside text for vision-based tasks like describing pictures, reading text from screenshots, and analyzing charts. The model supports tool use, reasoning, and caching features.
v0 apps can now securely connect to over 100 third-party services including Slack, Google, Notion, GitHub, and Salesforce through Vercel Connect. Team-managed connectors support both provider-handled registration and custom credentials, with short-lived tokens eliminating the need to store or rotate long-lived secrets.
OpenAI lowered list pricing for GPT-5.6 Sol, and Vercel's 50% AI Gateway discount continues to apply to the new lower rates through September 18th. Input tokens are 20% cheaper and output tokens are 33% cheaper across all service tiers with no code changes required.
Always-on tracing continuously collects traces from production and preview traffic, allowing you to debug real user requests without reproducing them. Control trace collection with sampling rules scoped to environments and paths, pay only for what you trace, and view traces in Logs or via the CLI.
You can now manage Vercel Toolbar comments directly from the command line using the Vercel CLI. The new vercel comments command covers the full triage loop, allowing you to list unresolved comments, inspect threads, reply, resolve, and edit comments from the terminal.
Developers can now emit custom metrics directly from Vercel Functions using the metric() function and analyze them alongside built-in observability data. Custom metrics can be queried, aggregated, and visualized in dashboards, and are available on Pro and Enterprise plans with Observability Plus.
v0 now uses a Snowflake request proxy for secure authentication in sandboxes, ensuring generated code cannot access or leak user OAuth tokens. The proxy resolves credentials at request time server-side, prevents credential injection into generated applications, and uses placeholder tokens instead of real credentials inside the sandbox.
Vercel Agent is now integrated with Slack code channels, allowing teams to collaborate with an AI coding agent directly in dedicated channels. The agent can follow work, accept instructions, review code and pull requests, and execute changes with approval workflows, all while maintaining audit trails and respecting user permissions.
Vercel Functions now support Bun 1.4, a major rewrite from Zig to Rust with over 2,900 resolved issues, improved Node.js compatibility, and new features. Update requires explicitly setting bunVersion to 1.4.x in vercel.json and reviewing breaking changes.
Vercel CLI now includes a new vcr command group to manage Vercel Container Registry. You can authenticate container tools like Docker and Podman with short-lived project-scoped tokens, then build and push images, create repositories automatically, and inspect images and tags using standard commands.
Algolia, a hosted search platform, is now available in the Vercel Marketplace for seamless provisioning and management. Installation automatically adds credentials to environment variables and includes agent skills for AI-powered coding integrations.
The Python SDK for Vercel Queues is now available in beta, enabling first-class Python support for background workloads with automatic retries, sharding, and delivery guarantees. Messages can be published and consumed from both JavaScript and Python runtimes within a single project.
Vercel Connect now includes a managed connector for Microsoft products like Teams and OneDrive. The connector handles Entra application registration and federated credentials automatically, allowing apps and agents to securely access Microsoft services with short-lived, automatically refreshed tokens scoped to requested permissions.
Fish Audio's text-to-speech and speech-to-text models are now available on the Vercel AI Gateway. All four models are free through September 18, after which standard pricing applies. Users can generate spoken audio from text, transcribe recordings, and access timestamped word-level details through the AI SDK or browser playground.
Vercel Agent is now available in Slack, allowing teams to investigate incidents, review code with production context, and take approved actions on deployments and configurations. The agent is read-only by default, respects user permissions, and operates with full auditability.
Chat SDK now includes an Instagram adapter, enabling bot development for Instagram DMs with support for media, quick replies, typing indicators, and reactions through Meta's Instagram Messaging API.
Exa's neural search engine is now available as a native integration on Vercel AI Gateway and Agent Marketplace, enabling models to search the web with a built-in tool requiring no separate Exa account, and allowing developers to deploy search and research agents directly from the marketplace.
Vercel Agent now integrates directly into Slack, allowing teams to mention it in channels and threads to get real-time diagnostics, infrastructure insights, and proposed fixes without leaving their conversation. The agent can investigate logs and metrics, identify root causes, and execute approved changes—keeping the full context of team discussions and decision-making in Slack.
A new Compliance section in Team settings lets you preview and download Vercel's compliance documents including SOC 2 Type 2 attestations and security policies. You can download individual documents, entire attestations, or multiple documents as a zip archive.
Vercel for Platforms now supports deploying directly from users' GitHub repositories using a GitHub access token, eliminating the need for users to install the Vercel GitHub App. Tokens are encrypted during retrieval and never stored on the deployment.
Vercel is launching a two-week public bug bounty program offering up to $1 million USD to security researchers who can demonstrate sandbox escapes in Vercel Sandbox. The program runs from August 18 to September 1, 2026, and rewards findings that break the Firecracker microVM compute boundary or defeat the network firewall controls protecting isolated workloads.
Sandbox CLI 4.0.1 improves output and error messages with actionable next steps, better formatting for create and fork commands, quieter errors with debugging support, an interactive sh shell, automatic update checks, and client-side defaults to ensure outdated installs don't break functionality.
GLM 5.3 from Z.ai is now available through Vercel's AI Gateway. The model offers improved performance on complex software engineering and multi-step agent tasks while producing fewer output tokens, and includes enhanced vulnerability discovery for security-focused applications. It maintains a 1M token context window and 128K max output, and supports function calling, structured output, streaming, and context caching.
Cline coding-agent runtime is now supported through the unified AI SDK harness layer interface, allowing developers to run Cline alongside other agents like Claude Code, Pi, and Grok Build without changing application code. Cline runs in the host process with sandbox integration for filesystem and shell operations.
Vercel KMS enables signing JWTs and messages from Vercel Functions using managed asymmetric keys, keeping private keys secure in the key management service instead of code or environment variables. Public keys are published via standard OIDC discovery and JWKS endpoints for verification anywhere.
Pro customers can now connect Cursor Origin repositories to Vercel, enabling automatic Preview deployments on pull requests and Production deployments on merge. Integration is available through Team settings, Project settings, or during project creation.
OpenAI's GPT-5.6 Sol model is discounted 50% on Vercel's AI Gateway through September 18, applying to all token types, tiers, regions, and modes. The discount covers cached tokens, cache writes, long-context requests, and works across Default, Flex, and Priority tiers with no code changes required.
Vercel CDN now supports Encrypted Client Hello (ECH) for Vercel DNS-managed domains, encrypting the Server Name Indication in TLS handshakes to prevent network observers from seeing which hostname a client is connecting to. ECH is enabled automatically at the platform level for supported browsers.
Vercel shares how its intern cohort contributed to major products over four months, including CDN routing rules, v0 features, billing infrastructure, and AI systems. The interns shipped features used by millions and gained real ownership and autonomy working on production systems.
GLM 5.2, an open-weights coding model with 1M-token context, is now free for eve agents through August 27 when served by Blackbox on AI Gateway. After the promotion ends, the model remains available at standard provider rates.
The AI SDK harness layer now supports any Agent Client Protocol (ACP)-compatible harness through the new @ai-sdk/harness-acp package, enabling integration with harnesses that don't have direct adapters. This meta-adapter wraps the ACP protocol instead of specific runtimes, letting you build an adapter for any harness that ships an ACP-compatible package.
Grok Build is now integrated into the AI SDK harness layer, allowing developers to run Grok Build through a unified interface alongside other supported coding-agent runtimes without changing application code.
Google's Gemini 3.7 Flash model is now available on Vercel's AI Gateway with a 50% discount through December 31st, 2026. The model improves on software engineering and agentic tasks with better reliability and generates desktop and web application code from design mocks.
Exa's neural search engine is now available as a native integration in the Vercel Agent Marketplace, enabling developers to add intelligent search and research capabilities to their apps with a single API key and unified billing through Vercel.
Team owners and members can now upgrade all projects using Node.js 20 or earlier to Node.js 24 directly from the Vercel Dashboard with a single click. The upgrade updates project settings but requires manual updates if a version is also defined in package.json.
DeepSeek V4 Pro now runs on updated weights (dated 0813) through AI Gateway. Existing requests automatically use the new weights without code changes, and you can explicitly target the updated model by specifying the deepseek/deepseek-v4-pro-0813 identifier.
Exa web search is now available for free through August 31 on AI Gateway and Eve, with no separate API key required. The tool is now the default web search for Eve agents and works with any AI Gateway model, returning web results with domain and date filtering support.
Vercel AI Gateway now supports routing for 9 different coding agents including Claude Code, Codex, and OpenCode. Developers can set up all agents in one command with centralized observability, billing, and policy controls across 200+ models.
Vercel built an autonomous software factory for the AI SDK open-source project that uses specialized agents to classify issues, reproduce bugs, implement fixes, and manage pull requests. The factory currently authors 25-35% of merged PRs and closes 70-80% of issues while keeping humans in control of all final merges, using isolated sandboxes for security and evidence-based assessments to guide reviewer efficiency.
Grok 4.6 from SpaceX AI is now available on Vercel's AI Gateway with a 500K token context window, support for text and image inputs, and configurable reasoning levels. Use it by setting the model to xai/grok-4.6 in the AI SDK or coding agents.
All Vercel Pro teams can now claim a free domain for one year across six TLDs. After the first year, domains renew automatically at regular pricing.
Vercel migrated the state management for the build warm pool from Redis to DynamoDB to provide durable storage for billing mappings while maintaining production availability. The migration used phased rollouts with feature flags and comprehensive validation to safely transition traffic without downtime.
Vercel Connect now provides line-level visibility into the token lifecycle, showing who created tokens, which app or project used them, and their current status. The new Observability tab includes runtime events, correlation IDs for system matching, and activity filters, with retention periods based on plan level and options to forward events to custom webhooks on Pro and Enterprise plans.
Enterprise Managed Users (EMU) is now generally available, giving organizations full control over Vercel accounts on verified domains through their identity provider. Managed users authenticate via SAML SSO only, with automatic account provisioning and lifecycle management through Directory Sync.
Vercel Connect now allows you to set up and integrate 100+ services entirely through the CLI. Previously, you could only complete terminal setup for some services and had to use the dashboard for others. The CLI automatically pre-populates connector details like brand name and auth type, then guides you through credential setup.
Vercel shares insights on AI-enabled defensive cybersecurity, highlighting how frontier models like OpenAI's Sol 5.6 can be used today for vulnerability discovery and defense, and introduces deepsec, an open-source security harness for scanning large codebases before open-weight models catch up in capabilities.
Vercel's August 2026 AI Gateway Production Index shows DeepSeek surpassed Google as the second-largest AI model by token volume, while average cost per token fell 13.6% due to widespread adoption of cheaper open-weight and frontier models. Open-weight models now represent over a third of all tokens on the gateway at significantly lower rates.
LaunchDarkly feature flags integration is now available on the Vercel Marketplace, enabling developers to set up feature flags quickly without additional configuration. Users can sync flags into Global Config, run experiments with rollback capabilities, and manage flags via the Vercel Toolbar.
Vercel Sandbox now includes a host-based firewall that controls network egress and injects credentials at the boundary, preventing untrusted code from exfiltrating data, scanning internal networks, or misusing stored credentials. The firewall supports domain allowlists, CIDR policies, temporary connectivity windows, and transparent TLS inspection without decryption for authorized connections.
Vercel Sandbox now uses Vercel Managed Images (VMI), versioned open-source base images that replace deprecated runtimes. The new universal image defaults to Ubuntu with Node.js, Python, coding agents, and standard utilities preinstalled, with nightly releases and automatic security updates.
Vercel Functions now support Bun.serve as a native entrypoint, allowing you to deploy servers with WebSocket support as-is without framework wrapping. Servers use Fluid compute with per-message pricing for WebSocket connections.
The deepsec open-source security review tool now offers one-command setup and improved initialization that automatically creates the isolated workspace, configures model access, generates codebase analysis, and runs initial scans. Setup is checkpointed at each step so interrupted runs can resume from where they left off.
Grok Imagine Image 2.0 Preview from xAI is now available on Vercel's AI Gateway. The model excels at detailed instruction following, maintaining structure in complex layouts like infographics and posters, and supports image editing with consistent subject generation across repeated calls.
Hermes Agent can now use Vercel AI Gateway as its inference layer with access to 200+ models at no token markup, and run agent commands in isolated Vercel Sandbox microVMs. Both features integrate into the setup wizard with live model availability and pricing information, and Sandbox supports Node 24, Node 22, and Python 3.13 runtimes.
Vercel Container Registry now supports making repositories public, allowing anyone with a Vercel account to pull and use images while maintaining read-only access. Repository visibility can be managed from the project dashboard, CLI, or REST API.
Audit Log Drains now stream team audit events to Datadog, Splunk, and Panther in addition to custom HTTPS and Amazon S3 endpoints. This replaces the Custom SIEM Log Streaming feature and is available on Enterprise plans.
New Pro subscriptions now include one free domain for the first year across six popular TLDs, with 30 days to claim it at checkout or from the dashboard. Domains renew automatically at regular pricing after the first year.
Terminal coding agents like Claude Code, Codex, and OpenCode can now each run in isolated Vercel Sandboxes, orchestrated through a tmux-style manager that displays multiple agents in side-by-side panes. Projects are uploaded to the Sandbox and agents run there without affecting your local machine, enabling safe parallel execution of multiple agents.
Agent Plugins 1.0.0 is now publicly available as an open, vendor-neutral standard for extending AI agents with reusable skills and MCP servers. The standard defines a portable contract using a minimal plugin.json manifest, allowing plugin authors to package components once and work across multiple supporting clients including ChatGPT, Cursor, GitHub Copilot, and VS Code.
Chat SDK now supports human-in-the-loop approvals through a new requestApproval method that posts a card with Approve and Deny buttons and suspends workflow execution until a decision is made. The wait can last any duration and survives deploys and restarts, with support for scoped approvers, verified decisions, and an audit trail.
Agent Plugins 1.0.0 is now available as an open, vendor-neutral standard for packaging Agent Skills and MCP servers into portable plugins that compatible agent clients can discover and load. The standard defines a common manifest format and fixed component locations, giving each client full control over installation, distribution, and user experience.
When installing integrations from the Vercel CLI, provider agent skills are now automatically installed alongside the integration, enabling agents to use the integrated tools without manual configuration. Users can also discover and browse available integrations directly from the terminal.
Seedance 2.5, a ByteDance video generation model, is now available on AI Gateway. It can generate up to 30 seconds of video with continuous camera movement and supports text, video, image, and audio inputs in a single request for more flexible creative workflows.
Ling 3.0 Tiny, a lightweight MOE model with 1.3B active parameters, is now available on Vercel's AI Gateway for free use through August 14. The model supports 256K token context windows, 32K output tokens, and native function calling, making it ideal for responsive agents and multi-turn conversations.
Users can now manually set custom avatars for their projects in the Vercel dashboard, replacing the automatic favicon detection. Avatars display throughout the dashboard and on Vercel bot comments in GitHub, enabling distinct branding for microfrontends and backend projects without favicons.
After purchasing a domain on Vercel, users are taken to a dedicated setup page that tracks domain registration in real-time and provides direct actions to deploy a new project, connect an existing one, proxy or redirect an external site, or configure email with DNS presets. The page displays registration progress and unlocks setup actions as the domain becomes ready.
AI Gateway now produces OpenTelemetry traces for every request and Pro/Enterprise teams can export them via Vercel Drains to any OTLP/HTTP-compatible endpoint, with native integrations for Braintrust, Dash0, Kubiks, Sentry, and Statsig. Traces include full request lifecycle details like model routing, fallback attempts, token usage, latency, and attribution metadata, with sampling controls and per-trace pricing of $0.05 per 1,000 traces.
97 more changes from Vercel. Sign up to read the whole changelog.
Sign up freeGitHub or email — no card needed.