OpenAI Agent Builder is a visual canvas for building multi-step AI agent workflows, but OpenAI is deprecating it, and as of October 2026 it's scheduled to shut down on November 30, 2026. If you've built something in it, export the workflow as Agents SDK code now; if you haven't started, don't start a new build there.
That warning isn't tucked away in a forum thread. It sits at the top of OpenAI's own Agent Builder guide, and the OpenAI deprecations page carries the same entry. So this page does four jobs: what the tool is, what happens on the shutdown date, where your workflow can go next, and what I'd do if you're only now looking at it.
What OpenAI Agent Builder does until it closes
Agent Builder lets you design an agent workflow on a canvas instead of in code. You start from a template or a blank canvas, drag in nodes for each step, give each step typed inputs and outputs, and preview runs using live data. OpenAI describes a workflow as a combination of agents, tools and control-flow logic, and every connection between two nodes becomes a typed edge, so each step knows exactly what the next one expects.
Its guide breaks the job into three steps. You design the workflow. You publish it, which turns it into an object with an ID and versioning; each publish creates a new major version that acts as a snapshot, and the canvas autosaves as you go. Then you deploy it, either by passing the workflow ID into ChatKit, OpenAI's framework for embedding chat on your site, or by downloading the Agents SDK code and running it yourself. While you build, Preview lets you run the workflow, attach sample files and watch each node execute, and an Evaluate tab runs trace graders over a trace or a set of traces.
The shutdown timeline, date by date
The guide says existing users can keep using Agent Builder during the transition window, and that ChatKit remains available. The deprecations page fills in the dates. What most people miss is that two neighbouring tools were announced on the same day:
| Date | What OpenAI's deprecations page says |
|---|---|
| June 3, 2026 | Deprecation announced for Agent Builder, with a notice sent to developers using it. The Evals platform and reusable prompt objects were announced for deprecation the same day. |
| Oct 31, 2026 | Existing evals on the Evals platform become read-only. |
| Nov 30, 2026 | Agent Builder is scheduled to shut down. The Evals dashboard and API, and the v1/prompts API with reusable prompt objects, are scheduled to shut down too. |
| After that | ChatKit remains available. The migration guide covers the Agents SDK and ChatGPT Workspace Agents. |
So if your workflow also leans on saved prompts or on graders to check its answers, plan for those at the same time rather than finding out in December.
Agents SDK or ChatGPT Workspace Agents: the two replacements
OpenAI's migration guide offers two paths and describes them in one line each. The Agents SDK is best for building agents through code. ChatGPT Workspace Agents are best for building agents through natural language and sharing them with teams.
If you searched for a ChatGPT agent builder, Workspace Agents is the route OpenAI points you to. There's a condition, though: the migration guide says you need a ChatGPT Business, Enterprise or Edu workspace with access to workspace agents and permission to create agents. Check your plan and your admin settings before you count on it.
And if you're not committed to OpenAI at all, this is a good moment to look around. I've compared other AI agent builders separately, so you can see what else does the same job.
How to export your workflow before it goes
Both replacement routes start from the same export, so do this first, even if you plan to keep running the workflow until the end of November. OpenAI's steps:
- Open your workflow in Agent Builder.
- Select Code in the top navigation.
- Select Agents SDK in the code dialog.
- Choose TypeScript or Python, then copy the complete export and save it somewhere you control.
For the Agents SDK route, you paste that export into your own application, install and configure the matching Agents SDK, and test the workflow in your runtime before you deploy. For the Workspace Agents route, you create a workspace agent in ChatGPT, paste the export into the chat and use the prompt OpenAI gives:
The prompt from OpenAI's migration guide
Please help me convert this workflow into an agent:
Work with meWant AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Before you click Create, the guide asks you to review the generated instructions and capabilities, set up any apps, tools, skills, authentication and connection permissions, select Preview and test representative inputs from the original workflow, and compare what you see with what the old workflow did. Only create the agent once it matches.
What won't carry over cleanly
The migration guide is refreshingly honest here. The process doesn't convert your workflow graph, and it doesn't guarantee every behaviour transfers unchanged. Three limits are spelled out:
- Workflows with strong determinism at their core may not migrate faithfully to a workspace agent. In plain words, if your workflow depends on step two always following step one in exactly the same way, test that hard.
- Connected apps, authentication, publishing and permission settings need a separate review in ChatGPT.
- On the Agents SDK route, you're responsible for checking the runtime configuration, tools, authentication, permissions and deployment yourself.
The safety side doesn't go away either. The Agent Builder guide flags prompt injection and data leakage as risks of agent workflows, and the migration guide tells you to follow the same safety practices, especially when the agent can reach private data or take actions through connected tools.
What I'd do if you haven't started yet
My opinion, plainly: don't start a new production build on a product with a published shutdown date. I'd skip Agent Builder for anything new, even a quick prototype you hope to keep. If your team lives in ChatGPT on a Business or Enterprise workspace and wants to describe agents in plain English, I'd go straight to Workspace Agents. If a developer will own it, the Agents SDK is the sturdier home. If you're mid-project, export today and pick a route this week, while the old version still runs and you can compare the two side by side.
If you just want to learn how agents work before spending anything, my guide to building an AI agent as a beginner without paying is a gentler place to start than a tool that's closing.
If you're working out which agent platform to build on so you don't have to rebuild it in six months, that's the kind of AI enablement I help with. I run AI training and workshops for teams, and help businesses with AI enablement and workflow automation, so the busywork runs itself and people get time back. Most of that work is with SaaS marketing teams and small businesses. Here's how we can work together.
Frequently asked questions
Is OpenAI Agent Builder shutting down?
Yes. OpenAI's Agent Builder guide says the product is being deprecated and is scheduled to shut down on November 30, 2026. The deprecations page says developers were notified on June 3, 2026.
Can I still use Agent Builder right now?
Existing users can keep using it during the transition window, according to OpenAI's guide. As of October 2026 that window runs until the scheduled shutdown on November 30, 2026.
What replaces OpenAI Agent Builder?
OpenAI's migration guide offers two routes: the Agents SDK for building agents through code, and ChatGPT Workspace Agents for building agents through natural language and sharing them with teams.
Is ChatKit being shut down too?
No. Both the Agent Builder guide and the deprecations page say ChatKit remains available.
Will my workflow migrate unchanged?
Not guaranteed. The migration guide says the process doesn't convert your workflow graph or guarantee that every behaviour transfers unchanged, so preview and test it before you rely on it.
