Vercel Workflows

6/17/2026

Introduces 'eve', an open-source agent framework built on Vercel Workflows. eve provides a structured, directory-based approach to defining agents, including models, instructions, tools, skills, subagents, and channels. It leverages Vercel's existing infrastructure for durable execution (Workflow SDK), sandboxed compute (Vercel Sandbox, Docker), human-in-the-loop approvals, and secure tool connections (Vercel Connect). eve also integrates OpenTelemetry for tracing and evals for testing.

4/16/2026

This post announces the general availability of Vercel Workflows, a new programming model for durable execution. It details the architecture, which comprises an event log, Fluid compute for step execution, and Vercel Queues. The post highlights the programming model, where workflows and steps are defined using 'use workflow' and 'use step' directives in TypeScript and Python. It emphasizes the benefits for agents, including durable execution, tool calling, state management, and handling interruptions, as well as durable streams for persistent agent output. Security features like default encryption and audit trails are also detailed. The post also mentions the integration with AI SDK v7's WorkflowAgent and provides examples of its application in building agents and long-running workloads.

3/31/2026

This post details how FLORA integrated Vercel Workflows with the AI SDK's DurableAgent to build their FAUNA creative agent. It highlights the use of Fluid compute for cost-efficient, parallel, long-running image generations and the benefits of durable orchestration for persisting, retrying, and maintaining state across agent steps. The post emphasizes the shift from managing separate orchestration systems (like Temporal) to an integrated AI stack, allowing for faster product iteration and reduced infrastructure debates.

2/27/2026

This post details the application of Vercel Workflows in building the Community Guardian and c0 agents for community management. It explains how Vercel Workflows enables durable execution, allowing agents to check in periodically, sleep, and resume, thus optimizing resource usage. It also highlights the integration with AI Gateway and AI SDK for model switching and agent interaction, and the use of the Chat SDK for multi-platform agent development.

2/9/2026

This post details how Vercel's internal data agent (d0) leverages Vercel Workflows, Sandboxes, Fluid Compute, and AI Gateway for reliable and scalable agent operation. It highlights the use of AI Gateway for model routing and cost/accuracy balancing, Sandboxes for secure execution of autonomous actions, and Fluid Compute for handling unpredictable, long-running agent compute patterns. The post also mentions Vercel Agent's capabilities in reviewing pull requests and analyzing incidents, all running on the same primitives.

1/12/2026

Mux has integrated their @mux/ai SDK with Vercel's Workflow DevKit to enable durable AI video workflows. This integration allows developers to build complex, multi-step AI processes without managing complex infrastructure. The SDK leverages "use workflow" and "use step" directives for opt-in durability, providing automatic retries, state persistence, and observability when deployed in a Workflow DevKit environment, while remaining a standard Node.js package otherwise. This enables features like content moderation, chapter generation, and summarization for video processing.

12/9/2025

This post details the framework integration pattern for the Workflow DevKit (WDK), explaining the two-phase build-time and runtime process for generating and exposing workflow handlers. It highlights the use of SWC compiler plugins for client, step, and workflow transforms, and discusses how WDK integrates with file-based routing frameworks (Next.js, SvelteKit, Astro) and HTTP server frameworks (Express, Hono) via Nitro. It also covers Hot Module Replacement (HMR) for developer experience and the conversion of framework-specific request objects to standard Web Request API.

11/24/2025

This post introduces Workflow Builder, an open-source Next.js template that provides a visual editor, execution engine, and infrastructure for building custom workflow automation platforms and agents. It leverages the Workflow Development Kit (WDK) and includes features like AI-powered text-to-workflow generation, webhook triggers, and referencing outputs from previous steps. The template integrates with Vercel AI Gateway and other services like Resend, Linear, and Slack.

10/23/2025

Introduces the Workflow Development Kit (WDK), an open-source TypeScript framework that makes durability a language-level concept for async functions. WDK uses `use workflow` and `use step` directives to turn async functions into durable workflows that can pause for minutes or months, survive deployments and crashes, and resume deterministically. It compiles steps into isolated API Routes, records inputs/outputs for replayability, and supports pausing workflows with webhooks for external events. WDK is designed to run on any platform via 'Worlds', with a Vercel World using Framework-defined infrastructure and a reference Postgres World provided.

10/15/2025

This post details the integration of Vercel AI Cloud with Slack, enabling agents to operate within conversational interfaces. It highlights the development of a Slack Agent template using Bolt for JavaScript and Nitro, and the introduction of v0, a collaborative agent for building and deploying sites, apps, and agents directly from Slack conversations. This extends the capabilities of Vercel Workflows to conversational UIs and enterprise collaboration platforms.

10/22/2024

This post details how the BNP Paribas Open, in partnership with Work & Co, utilized Vercel's platform, including Next.js and serverless functions, to build a digital experience capable of delivering real-time scores and fan engagement. Key technical aspects highlighted include the platform's scalability and performance for handling millions of users during peak loads, rapid development and deployment capabilities enabling a tight four-month timeline, flexible deployment processes for quick rollouts and rollbacks, and an integrated ecosystem combining CDN, static pages, dynamic functions, and data services. The success of this project is attributed to Vercel's ability to manage complexities of scalability, deployment agility, and real-time performance, allowing the team to focus on building digital experiences. The post also mentions improvements in SEO due to server-side rendering and personalized content delivery.

2/22/2023

Introduces Vercel Cron Jobs, a feature for scheduling Vercel Functions. It details configuration via `vercel.json` with cron expressions, provides examples of schedules, and outlines limits for different plans. It also mentions the potential for framework-native integration via the Build Output API.