Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
ReleaseCode / SDK · ActionableX: @rauchg→ original

Vercel Open-Sources a Visual Builder for AI Agents and Workflows

Vercel has open-sourced the code for a visual builder where AI agents and workflows are assembled from blocks rather than written from scratch. The output is ordinary code with the "use workflow" directive, which can be taken into your own project. There is also a "text to workflow" mode: you describe the task in words, and the AI assembles the diagram. For teams that don't want to depend on closed no-code platforms, this offers visibility while keeping the code in their hands.

AI-processed from X: @rauchg; edited by Hamidun News
Vercel Open-Sources a Visual Builder for AI Agents and Workflows
Source: X: @rauchg. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceX: @rauchg ↗
Editor-in-Chief
Fully open sourcethe project can be cloned and deployed on your own
Key metric #2Built on Vercel's AI SDK and AI Elements UI components
Sample integrations includedResend, Linear, and Slack

Nearly every team building AI agents eventually runs into the same problem: the logic grows into a chain of steps, where one email, one model call, and one message in a work chat depend on each other. Keeping this in your head and in code is hard, and explaining it to a product manager or a colleague in another department is harder still. Guillermo Rauch, Vercel's CEO, announced a release that tries to address this pain: a visual builder for agents and workflows with fully open source code.

What Was Released

According to the post, this is a builder where a process is assembled as a diagram of blocks: a trigger, a model call, a notification, and so on. The main difference from familiar no-code services is that the result does not stay inside someone else's platform. The builder outputs code with the "use workflow" directive. This marker denotes long-running processes in the Vercel ecosystem: a function with this annotation can be multi-step and is designed to handle individual steps failing and needing to be retried.

In simple terms, you draw a diagram and get a regular file in your repository. You can open it in an editor, edit it by hand, run it through code review, and commit it to Git. For an engineering team, this is the key point: the picture helps agree on the logic, but the source of truth remains the code.

Another advertised feature is "text to workflow." You describe the task in plain language, for example "when a critical bug appears in Linear, post to Slack and send an email to the on-call engineer," and the model generates a draft diagram. It then needs to be checked and refined. The author of the post does not provide quality metrics for this generation, so you should not expect it to assemble flawlessly.

What It's Built From

The tech list names AI SDK and AI Elements. The first is Vercel's toolkit for working with models from TypeScript applications. The second is a set of ready-made interface components for chats and AI features. This suggests the builder is unlikely to be tied to a single model provider, since AI SDK exists precisely to swap models without rewriting an application. However, the post offers no details on which providers are supported.

The sample integrations include Resend for email, Linear for tasks, and Slack for messaging. This is a typical set for internal automation: send a notification, open a ticket, post to a channel. Given the phrasing "sample integrations," they seem to serve as examples. If you need a different service, you will have to write your own block by analogy.

What to Do With This, and What Remains Unknown

The practical takeaway for developers is this. If you plan to build internal automations or agents on a Next.js and Vercel stack, you can clone the repository and see how the blocks are structured and how a visual diagram turns into code. Even if the builder doesn't suit you, it is a ready-made learning example of how to combine agents, long-running processes, and an interface. For product managers and founders, the value lies elsewhere: a visual diagram helps discuss agent behavior without reading code, and the open license eases concerns about vendor lock-in.

Now the limitations. The source is a single post on X, and it contains few details. The license is not named, and it is not stated whether the project can run outside Vercel's infrastructure, how state storage and retries work, or how much it costs to execute processes in production. The "use workflow" directive is tied to its own runtime environment, so moving to another hosting provider may prove difficult. This should be verified before building critical business processes on the builder.

Looking more broadly, the release fits a noticeable trend: developer platforms are trying to offer not only models and SDKs but also a ready-made agent orchestration layer. Vercel is betting on open code and on keeping the output as regular code rather than configuration in someone else's cloud. How well this idea holds up under real workloads will become clear once the first reports from teams that put such workflows into production appear.

Story

Key Questions Answered

What Was Released?

According to the post, this is a builder where a process is assembled as a diagram of blocks: a trigger, a model call, a notification, and so on. The main difference from familiar no-code services is that the result does not stay inside someone else's platform. The builder outputs code with the "use workflow" directive. This marker denotes long-running processes in the Vercel ecosystem: a function with this annotation can be multi-step and is designed to handle individual steps failing and needing to be retried.

What It's Built From?

The tech list names AI SDK and AI Elements. The first is Vercel's toolkit for working with models from TypeScript applications. The second is a set of ready-made interface components for chats and AI features. This suggests the builder is unlikely to be tied to a single model provider, since AI SDK exists precisely to swap models without rewriting an application. However, the post offers no details on which providers are supported.

What to Do With This, and What Remains Unknown?

The practical takeaway for developers is this. If you plan to build internal automations or agents on a Next.js and Vercel stack, you can clone the repository and see how the blocks are structured and how a visual diagram turns into code. Even if the builder doesn't suit you, it is a ready-made learning example of how to combine agents, long-running processes, and an interface.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Autonomous AI Execution · Nomad Agent

Delegate Deep Research, Coding & Multi-Step Workflows to Nomad Agent

Deploy autonomous task execution with Nomad Agent (agent.nomadnet.ai) — automated web research, data synthesis, code generation, and MCP tool orchestration.

Launch Nomad Agent →Nomad AI Workspace
What do you think?
Loading comments…