Google Opal Is Closing on 17 November: Save Your Prompts and Choose What Comes Next

You have a small Opal app that saves you from typing the same instructions every week. It may turn notes into a quiz, prepare a content brief or build a report. Now the Google Opal shutdown means you need to keep that useful work before the service closes.

Collect your prompts, understand the steps behind them and decide what you actually need to rebuild. Start with one important workflow. You can make a sensible plan without moving every old experiment.

Confirmed by Google: Opal and Gems made by Labs turn off on 17 November 2026. Saved workflow files remain in My Drive → Opal, but the interactive workflows stop working. They will not move automatically to a replacement. Google recommends Gemini skills or AI Studio, depending on the task. Read Google’s shutdown FAQ.

Research note: Official guidance checked on 21 October 2026. The backup plan and example below are practical recommendations. The example is illustrative; this article does not claim an independently tested Opal export or completed account migration.

What the Google Opal shutdown means for your work

Think about your workflow in three parts: the instructions you wrote, the connections between its steps and the final experience someone uses. Saving the instructions gives you a useful starting point. You should still review how the new version will collect inputs, run tools and present the result.

Google describes Opal as a visual workflow system with input, generation and output steps. A prompt can refer to an earlier step, a tool or an asset. Those references explain why copying one text box may leave out important parts of the app. Google’s editor overview explains these parts.

For example, a revision-quiz app might accept a chapter, create questions and display them on a webpage. When planning its next version, check the chapter input, question-writing rules and display separately. That makes the work easier to estimate than a broad request to “move the app”.

Do ordinary Gemini Gems migrate automatically?

There are two different products behind similar names. Google says ordinary Gems will be moved into skills when Gems are retired. Its skills announcement specifically excludes Gems by Google Labs from that automatic migration. The Labs mini-apps are the Opal-powered workflows discussed here. Check the announcement and its second footnote.

If your work opens as a visual app with connected steps, check its identity before relying on advice about ordinary Gems. Keep a note of the app name, owning account and editor link. A screenshot of the editor can also help you remember what you are preserving.

Before closure: how to save Opal workflows

Use this checklist for the workflows you still need. These are recommended backup steps; the filenames below are your own organisational choices.

1. Make a short list of the workflows worth keeping

Write down the name, purpose and owner of each useful app. Add a priority: used regularly, useful occasionally or no longer needed. Start with the app that would interrupt real work if you lost access to it.

For each important workflow, answer three questions:

• What does the person provide at the start?
• What should the finished result contain?
• Which part takes extra work outside the prompt?

That last question could uncover a web lookup, an image-generation step or a spreadsheet output. Keep it visible in your notes so it does not disappear during the rebuild.

2. Find the matching workflow file in Drive

Using the workflow’s owning account, look in My Drive → Opal, the folder named in Google’s FAQ. Match the file to the app you are documenting.

On a computer, select a file in Drive, right-click and choose Download. You can select several files with Ctrl on Windows or Command on a Mac before downloading. If download is unavailable on a shared file, check with its owner; Drive can restrict downloading for viewers and commenters. Google documents these download steps and restrictions.

Keep the downloaded original unchanged. Make another copy if you want to inspect or annotate it. A useful local folder name is opal-backup-2026-10-21. Add the date you actually make your backup.

3. Copy prompts and record what each step uses

Open the workflow in the editor and select each relevant step. Google’s documentation says the sidebar exposes a selected step’s prompt. Check references to other steps, tools and assets as you copy it. See the editor’s step-details guidance.

In a plain-text note, record the following beside each prompt:

• Step name and purpose: what it is meant to do.
• Input: a user message, uploaded material or an earlier result.
• References: the files, tools or previous steps it depends on.
• Expected output: the content and format you need.
• Model or settings: anything visible that may affect the result.

A reference such as “use the previous step” will be hard to understand later unless you also record what that previous step produced. Give it a clear description now.

4. Keep a small example and the useful output

If you can still run the app, use a short input that you own and save the output you consider acceptable. Note what makes it useful: perhaps the number of questions, the tone or the way answers appear.

Save important images, documents and other assets separately where you have permission. Check that you can open those copies. Avoid assuming a workflow file contains every external asset or every output you may want later.

For a revision quiz, one short paragraph and its checked answer key are enough to begin. You do not need a large collection of private student or customer material to test the first version.

Choose between Gemini skills and Google AI Studio

For Google Opal migration, choose around the job you want to keep doing. Our recommendation is to rebuild the smallest useful version first, then add any missing connections.

These are starting points for evaluating Opal alternatives. They are not a compatibility chart or a promise that the original app will behave the same way.

Use Gemini skills for instructions you repeat

Skills save reusable instructions inside Gemini. You can call a skill in chat or allow Gemini to use a relevant activated skill. The current help page also says skills can be used in chats without a Google AI subscription. Check the current skills guide for access and requirements.

For a simple text-only rebuild on the web:

1. Open Gemini and go to Customize → Skills. If Customize is hidden, check More.
2. Choose Create → Create manually.
3. Enter a focused name, description and your saved instructions.
4. Create the skill, then select it in a new chat using /, or @ if your account has the updated shortcut.

These labels follow Google’s current creation guide. Menu labels can differ during rollout. If you do not see the feature, check the account’s availability instead of assuming your backup failed.

Do not assume that every former tool is supported. Google’s transition help currently lists limits involving features such as video creation, Canvas and Deep Research. Check the exact capability your old app depended on before choosing this route. Review the current tool limitations.

Use AI Studio for a closer look at the prompt

AI Studio is useful when you want to experiment with prompts and model settings. Google’s quickstart documents a Playground, system instructions and a Run settings panel. It also provides code export for taking a prompt towards an API implementation. Follow Google’s AI Studio quickstart.

1. Open a new chat prompt in AI Studio.
2. Put the reusable task rules in System Instructions, available through Run settings.
3. Put your example input in the message field.
4. Choose an available model and record the settings.
5. Run the prompt and check it against your saved example.

Begin with a single stage. If you combine several old stages into one prompt, test whether anything important gets skipped. Keep the task rules separate from the changing input so later corrections are easier to follow.

Our Google AI Studio overview explains the platform’s main purpose. For current model access and billing, use Google’s documentation linked here.

A small rebuild example: turn notes into a revision quiz

Illustrative example: this is a new prompt you can try. It is not an exported Opal file or a reported test result.

Suppose your old app accepted study notes and returned multiple-choice questions. First rebuild the question-writing part. Give the skill a name such as create-revision-quiz and use this description:

Creates a short revision quiz from supplied study notes. Use when the user asks for questions based only on the material they provide.

Use the following instructions in a manually created skill, or as system instructions in AI Studio:

Create a revision quiz using only the notes the user supplies.

1. Write three multiple-choice questions.
2. Give four options for each question, labelled A to D.
3. Make sure each question has one correct answer.
4. Put the answer key after all three questions.
5. Add one short explanation for each answer.
6. Use simple English. Do not add facts from outside the notes.
7. If the notes cannot support three clear questions, explain
   what is missing instead of inventing information.

Try this short input:

Notes:
Plants use light energy to make food through photosynthesis.
Chlorophyll absorbs light.
Carbon dioxide and water are used in the process.
Oxygen is released.

Create a revision quiz from these notes.

Now check the output yourself. Are there exactly three questions? Does each have one correct answer? Does the answer key match the options? Has it added a fact that the notes never gave?

Then try an incomplete input, such as “Plants need light.” The instruction should lead it to explain the missing material when it cannot make three useful questions. Record what actually happens; a prompt is an instruction, not a guarantee.

The design follows Google’s advice to keep a skill focused, describe when to use it, give a clear output format and explain how to handle missing information. Read Google’s skill-writing guidance.

If the old mini-app displayed questions on a custom webpage or saved them elsewhere, add those requirements to a separate rebuild list. The text-generation test does not establish that the interface or external actions have been recreated.

Check access and costs before rebuilding everything

Your Google Opal shutdown plan should include a small cost check. Run one representative task, note the chosen product and model, and check how its allowance is measured before estimating a larger workload.

Google lists free and paid Gemini API access, with model-specific availability and charges. A free starting route does not mean every model or tool is free. Moving from a prompt experiment to an API-based app can introduce usage charges. Check the live Gemini API pricing page.

Write down the account, model, input size, number of attempts and any connected services used. For an important workflow, also count the time spent checking the output. Avoid choosing a replacement on a headline price alone.

After closure: retrieve the prompt text from your file

For the Google Opal shutdown, Google says you can still download saved workflow files from Drive after closure and open them in Notepad, TextEdit or another plain-text editor to retrieve prompts. That retrieves text; it does not restart the hosted app. This recovery route is documented in the FAQ.

Work on a copy. Search for a distinctive sentence you remember, then inspect the nearby text. A structured workflow file may include settings and references around the prompt. Check which material is an instruction and which part is only a reference to another step.

If the file is missing, check the original Google account, your backup location and the owner of any shared workflow. Do not assume there is a universal recovery tool. Confirm what you can actually retrieve before planning the rebuild.

Common questions before choosing the next version

Does renaming a downloaded file make it a Gemini skill?

A filename change does not explain how the old inputs, instructions and outputs should work in the new product. Use the supported creation process and review the result. Avoid treating a renamed file as a completed migration.

Should I put every old step into one long prompt?

Start with one complete, useful task. If it fails, separate the stages and check the input and output at each point. A shorter working version is easier to improve than a long prompt whose failures you cannot locate.

How do I know the new version is ready?

Check a normal input, an incomplete input and one that previously caused trouble. Review the output, permissions and any external actions. Keep the replacement in a trial stage until it does the job you need with acceptable results.

Your final save-and-rebuild checklist

With Opal closing, the most useful next step is a small verified backup. Use this list to finish the work:

• Identify the owner and location of each important workflow.
• Keep an unchanged copy of its saved file.
• Copy the prompts and explain their references.
• Save essential assets and a checked example output.
• Choose a replacement around the task, not the product name.
• Rebuild one useful part and record the missing pieces.
• Test normal and incomplete inputs before depending on it.
• Update any shared instructions or links after checking the replacement.

You do not need to rebuild every experiment. Preserve the work you would miss, make its instructions understandable and confirm that the next version produces a result you can use.