Thanks, but I still have this question. My clients use assistants available from the dashboard and I give them the opportunity to use my app via Assistants API. I can see assistants in my dashboard too and don’t see any indications they are going to go, no any banner or something like that, neither do my clients. I’m talking about access to all the existing assistants. The API is deprecated, will assistants be shut down too? If so, where can I read about it? Thanks.
@KathSeaCat If you edit or create a new assistant from the dashboard, there will be a “Learn more” button with this link.
If you are the organization owner, you probably also received this email:
This is about Assistants API, I’m asking about assistances in the OpenAI dashboard. From the assistants dashboard there is literally nothing indicating they will be shut down. My clients work with the assistance and don’t use or read the OpenAI docs. So, if OpenAI is going to deprecate the way how I’m working with my clients’ assistants, should they provide an adequate way to keep doing it, OR will the assistants themselves be gone?
This is my dashboard:
The dashboard is just a frontend based on the API.
And the link is there in the button I showed you.
Click “edit” or “create” in the same screen and you shall see the “Learn more” button.
Okay, but it’s still is super unclear as it says:
- In the dashboard, recreate that bundle as a named prompt.
What is that, where can I find them? My dashboard doesn’t have any “named prompts”.
In the dashboard over the “Chat” menu, you can create reusable prompts. You can read more here:
It is quite an extensive subject, I recommend scheduling some time before the deprecation final date to study it further.
I know, it is a lot to take. Wish you luck with your migration.
The platform site is NOT an end-user application. It is a place to make trial use of an AI model and develop an idea. Hence, the name “Playground”.
Why would people say you should develop your own application and not have “clients” consume your API organization directly through the portal? Because API means “application programming interface”, a standardized way to connect two computer programs together, a server (OpenAI) to a client (software you make).
That does not change that not only is assistants deprecated, but OpenAI could take away the Platform site demonstrator at any time, just like they did for the completions endpoint playground, although the endpoint still is operational, and just like they completely broke the Chat Completions playground preset creation and editing purposefully.
If you want to share a customization with instructions and some files to search, and are an AI product consumer and not really an AI product developer, I would look at GPTs in ChatGPT Plus or higher. They are designed to be shared with others, are based on the same general architecture… and the monthly subscription fee per user goes directly to OpenAI.
Ahh okay, I finally see it. Honestly, it’s all a total mess. Chat is actually chat prompts which are actually an assistants replacement. Oh, my. It’s such a pain. Thanks.
Totally agree! The naming really confuses people at first. I’ve been using the Assistants API for a while, and yeah it always felt kind of beta. Definitely planning to migrate stuff sooner rather than later before it gets fully shut down.
In the recent migration guide (https://platform.openai.com/docs/assistants/migration?user-chat-app=responses#migrating-your-integration) it is mentioned that we should click the button “Create prompt” to easily migrate the assistants to Prompts but I don’t see that button anywhere in my dashboard. How can I access that?
Sounds like “create prompts” as pictured was made up by an imaginative AI. “This will create a prompt object out of each existing assistant” implies there is some method for selecting them that does not exist.
I can identify no surface for assistants, either in listing, using, nor editing that offers any kind of export. Responses’ “Prompts” is made especially useless because there is no programmatic API creation of them, where one could actually write a migration tool. Probably intentional: “developers can no longer make an analogue of user-created GPTs”.
You can just copy instructions and vector store IDs and the text of functions. The most tedious would be if you use multiple preset code interpreter file IDs, and the need for containers to be created that you automatically start paying for.
I have another issue regarding feature parity. We currently use assistants in our production environment - allowing users to upload a Word Document (.docx) for the ai assistant to open and summarise.
I add the file using Files API and pass the file_id to the assistant / response api.
I have converted to using prompts and conversations, however it appears that I cannot now add a docx, and must convert it myself to pdf first. The prompt even returns advice on how to do this when I questoin it about the issue and acknowledges that there is not feature patity here.
Can we please have this ability restored?
This is a big milestone 
. The Assistants API played a key role in exploring what agent-like systems could become, and it’s great to see its strengths carried forward into the Responses API. The addition of built-in tools, persistent conversations, and reasoning token continuity with GPT-5 feels like a huge leap forward for developers. The migration guide is also a thoughtful touch — making it easier for teams to adapt without friction. Excited to see how the Responses API sets the new standard for advanced multi-step workflows and seamless integrations! 
Hello,
I’m not sure if this is the best place to mention it, but the link to the Assistant Migration guide in this paragraph is incorrect.
It currently points to:
https://platform.openai.com/docs/guides/assistants/migration
The correct link (found on the first message of this conversation) is:
https://platform.openai.com/docs/assistants/migration