> ## Documentation Index > Fetch the complete documentation index at: https://docs.typingmind.com/llms.txt > Use this file to discover all available pages before exploring further. # Custom GPTs Shutdown (Dec 11, 2026): How to Migrate Your GPTs > OpenAI Custom GPTs stop running on December 11, 2026. Back up your GPTs and rebuild them as TypingMind AI agents with the same instructions, files and tools. > [!WARNING] > > **Custom GPTs stop running on December 11, 2026.** Back up your GPTs' instructions and files before then — after that date their settings can't be opened. OpenAI is retiring **Custom GPTs**, and they stop running on **December 11, 2026**. If your writing, support, or research workflows run on them, they need a new home before then. This guide covers what changes, what you'll lose, and how to rebuild your GPTs as **TypingMind AI Agents** that work with any model. [YouTube video player](https://www.youtube.com/embed/mNIsgxk8Dps) ## What's changing On September 11, 2026, OpenAI announced it will retire Custom GPTs on every ChatGPT plan and replace them with **ChatGPT Plugins**. | Date | What happens | | :- | :- | | Sep 11 | Retirement announced | | Oct 26 | New GPT creation ended in affected Enterprise workspaces | | Dec 11 | Custom GPTs stop running | | Feb 11, 2027 | Custom GPTs stop running in FedRAMP workspaces and Enterprise workspaces with an approved deferral | Personal plans (Free, Go, Plus, Pro) already can't create new GPTs. Timing can vary by workspace, so check your account notifications. OpenAI's built-in migration only moves part of a GPT: | Moves to a plugin | Doesn't move | | :- | :- | | Instructions (as a skill) | Custom Actions | | Knowledge files | Your chosen model | | Connected apps | Sharing settings and users | Your past conversations with Custom GPTs stay in your ChatGPT history after retirement, but they can't be moved into a ChatGPT project. Once migrated, the original GPT becomes read-only. If your GPT does more than instructions and files, you're rebuilding part of it either way, so it's worth choosing where. ## Why TypingMind AI Agents AI Agents are TypingMind's equivalent of Custom GPTs: reusable assistants with their own instructions, knowledge, and tools, without being tied to one provider. > [card] Any model, per agent > > Run each agent on GPT, Claude, Gemini, DeepSeek, Mistral, or a custom endpoint, with its own parameters. > [card] Knowledge & RAG > > Attach files, connect a Knowledge Base, or pull live data from your own API. > [card] Actions as plugins or MCP > > Rebuild Custom Actions as HTTP or JavaScript plugins, or connect MCP servers. > [card] Built for teams (available via Teams solution) > > Share agents by link, control access by user group, and roll back any change. | Custom GPT | TypingMind | | :- | :- | | Instructions | Agent system instructions | | Conversation starters | Conversation starters and welcome message | | Knowledge files | Training files or Knowledge Base | | Custom Actions | Plugins or MCP servers | | Web, image, Code Interpreter | Search plugins, image plugins, Code Sandbox | | Selected model | Any model you assign | ## Rebuild a GPT in TypingMind Repeat these steps for each GPT you want to keep. Steps 5 and 6 only apply if your GPT used built-in tools or Custom Actions. ### Step 1: Back up your GPT Do this first, while the GPT still opens. After December 11 its settings are gone. 1. In ChatGPT, open **My GPTs**, find the GPT, and click **Edit**. 2. On the **Configure** tab, copy these into a document: * Name and description * Instructions * Conversation starters * Which capabilities are turned on (web search, image generation, code interpreter) * The recommended model, if one is set 3. Find the original copies of your **knowledge files**. 4. If the GPT has **Actions**, open each one and save the full schema, the authentication type (None, API Key, or OAuth) and its settings, and the privacy policy URL. 5. Export your chat history from **Settings → Data Controls → Export Data**. OpenAI emails you a download link. > [!WARNING] > > You may not be able to view saved API keys or OAuth client secrets in the action editor. Make sure you have them elsewhere, or can generate new ones with the API provider. ### Step 2: Set up TypingMind 1. Get API keys from the providers you plan to use, such as OpenAI, Anthropic, or Google. 2. In TypingMind, open **Models** in the left sidebar, switch to the **API key** tab, and add each key. 3. Import your chat history from **App Settings → All Data → Import From OpenAI**, then select the export file from step 1. > [!NOTE] > > You pay each provider directly for what you use. You don't need ChatGPT Plus, and it doesn't cover API usage anyway. ### Step 3: Create the agent 1. Click **Agents** in the left sidebar, then **Create AI Agent**. 2. Enter the name and description from your backup. 3. Paste the instructions into the system instructions field. 4. Under **Base model**, toggle **Model & parameters** and choose a model. Start with the one your GPT used, or try another provider to compare. 5. Add your **conversation starters**, and optionally a **welcome message** the agent sends first. ![Ai Agent](https://mintcdn.com/typingmind/BH2e7op4JnHUepwV/images/ai-agent-1.webp?fit=max\&auto=format\&n=BH2e7op4JnHUepwV\&q=85\&s=710aaa2fe94b22d492ac8fdb9f5e97be) > [!TIP] > > If your GPT had to follow a strict output format, add a few example exchanges as **few-shot prompts**. They're inserted after the system instructions in every chat. ### Step 4: Add knowledge Choose one option under **Knowledge** in the agent editor, based on how much material the agent needs: | Option | Best for | How it works | | :- | :- | :- | | **Training files** | A few short documents | Files are added to the system prompt. Assign a model first, and keep files within its context window. | | [**Knowledge Base**](https://docs.typingmind.com/rag-knowledge-base.md) | Many or long documents | Upload documents in the **KB** menu, add training tags, then connect those tags to the agent. Only relevant passages are retrieved for each question. | | [**Dynamic Context via API**](https://docs.typingmind.com/ai-agents/dynamic-context-via-api.md) | Data that changes often | Live content from your own API or vector store is added to the prompt. | ![Kb](https://mintcdn.com/typingmind/8OOIsQ9c2aqWN9qQ/images/kb.webp?fit=max\&auto=format\&n=8OOIsQ9c2aqWN9qQ\&q=85\&s=0688ce167a1ac5928590186af5bd301e) ### Step 5: Turn on tools Replace each capability your GPT used: | GPT capability | In TypingMind | | :- | :- | | Web search | The model's built-in web tool, or a search plugin such as Perplexity or SerpAPI | | Image generation | GPT Image Editor or Stable Diffusion plugin | | Code Interpreter and data analysis | Code Sandbox | ![Tools](https://mintcdn.com/typingmind/AIWS1xg988vqSxYj/images/tools.webp?fit=max\&auto=format\&n=AIWS1xg988vqSxYj\&q=85\&s=d144b7524712554a18a44c075ba11af4) 1. Open **Plugins** in the left sidebar, install the plugins you need, and fill in any settings they ask for. 2. In the agent editor, toggle **Plugins** under **Skills** and select them. ### Step 6: Rebuild Custom Actions as plugins Each Custom Action becomes a **TypingMind plugin**. Every operation in the action's OpenAPI schema turns into one plugin function that calls the same API. | In your GPT Action | In a TypingMind plugin | | :- | :- | | Each operation (`operationId`) | One function, defined by an **OpenAI Function Spec** | | Operation summary and parameters | The function's `description` and `parameters` | | Server URL, path, and method | The **HTTP Action** method and endpoint URL | | API key authentication | A **User Settings** field, used as `{apiKey}` in a header | | OAuth | The plugin's **OAuth 2.0** settings, used as `{OAUTH_PLUGIN_ACCESS_TOKEN}` | | Response | The **output**: give it to the AI, or render it as Markdown or HTML | Learn more about how to [build custom plugins on TypingMind](https://docs.typingmind.com/plugins/build-a-typingmind-plugin.md) ### Step 7: Test side by side Finish this before December 11, while the original GPT still runs. 1. Collect a handful of real prompts you've used with the GPT, including ones that trigger its knowledge and actions. 2. Run each prompt in the old GPT and the new agent. Compare tone, format, accuracy, and whether the right tools were called. 3. Adjust the instructions, model, or parameters until the agent matches or beats the original. Agent version history lets you roll back any change. > [!TIP] > > Bring your chat history to TypingMind: **App Settings → All Data → Import From OpenAI** imports your ChatGPT export. > > ![Import Openai 1](https://mintcdn.com/typingmind/8OOIsQ9c2aqWN9qQ/images/import-openai-1.webp?fit=max\&auto=format\&n=8OOIsQ9c2aqWN9qQ\&q=85\&s=ffb279e600168851943a7e94bdd11c71) ## FAQ > [accordion] Can I import a Custom GPT automatically? > > Not the GPT itself. You rebuild it as an agent from your backup, which usually takes a few minutes. Your ChatGPT conversations can be imported automatically. > [accordion] Do I still need ChatGPT Plus? > > No. TypingMind uses your own API keys, billed by each provider. ChatGPT Plus doesn't cover API usage anyway. > [accordion] Where is my data stored? > > Chats and API keys stay in your browser by default, and requests go straight to the provider you choose. Cloud Sync is optional. ## Related guides > [card] [Migrate from ChatGPT](https://docs.typingmind.com/migrate-from-chat-gpt.md) > > Map every ChatGPT feature to TypingMind. > [card] [AI Agents overview](https://docs.typingmind.com/ai-agents/ai-agents-overview.md) > > Everything you can configure in an agent. > [!NOTE] > > Dates reflect OpenAI's [Custom GPT retirement FAQ](https://help.openai.com/en/articles/20001519-custom-gpt-retirement-and-migration-faq) as of October 1, 2026 and may change. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.