GitHub Spark | GitHub.com Experience Retires August 31

GitHub is retiring the current GitHub Spark experience on github.com after shifting its application-building strategy toward more integrated Copilot workflows. Beginning August 4, 2026, Spark stopped accepting new users and no longer allows new apps to be created, while existing users have until August 31 to access their projects and export the code they want to continue developing.


GitHub Spark retirement and app export deadline on github.com

{getToc} $title={Table of Contents}

GitHub is moving app building toward integrated Copilot workflows


GitHub Spark was designed to reduce the distance between an idea and a working full-stack application. Users could describe an app in natural language, refine it through visual tools or code, preview the result, and deploy it through a managed environment without assembling the complete development stack manually.


GitHub is now positioning environments such as Visual Studio Code, Copilot CLI, and the GitHub Copilot app as the preferred places for building and refining these experiences. The company specifically describes the retirement as applying to the current Spark experience hosted on github.com, rather than presenting the change as a general retirement of AI-assisted application development across GitHub.



Existing Spark projects should be exported before August 31


Existing users can continue accessing apps they already created until August 31, 2026. Anyone who wants to keep editing a Spark project after that date should export its code by opening the application's Spark workbench, selecting the additional options menu, and choosing Create repository.


This step moves the application code into a regular GitHub repository, where development can continue using standard GitHub workflows. Spark already supported repository integration, Codespaces, pull requests, issues, GitHub Actions, and Copilot, so exporting the project creates a path from the retiring interface into the broader GitHub development environment.


Already deployed Spark apps will remain available


The retirement does not automatically take deployed Spark applications offline. GitHub says apps that were already deployed will continue to work after the Spark experience shuts down, which gives existing users some separation between preserving a live application and preserving the ability to keep developing it.


That distinction is important for creators who used Spark for prototypes, internal tools, interactive web experiences, or small production applications. A deployed project may continue serving users, but exporting its code before the deadline is necessary if the creator wants a conventional repository that can be maintained and extended later.


Apps using llm() need a new inference provider


AI-powered Spark applications need an additional compatibility check. GitHub Models, which supplied inference for Spark's llm() function, was retired on July 30, 2026. As a result, calls to llm() no longer work even before the August 31 Spark retirement date.


Creators can search their application code for llm() to determine whether the project is affected. If those calls are present, the application must be connected to another inference provider to restore its AI functionality. Users will also need to provide their own API credentials and manage the associated inference billing themselves.


IMPORTANT: Existing GitHub Spark users should export app code before August 31, 2026 if they want to continue editing their projects. Apps using llm() also need a separate inference provider because the GitHub Models service previously used by that function has already been retired.{alertWarning}

Daisuki's Take: What This Means for Web Designers


For web creators, the most important part of this announcement is the migration path rather than the shutdown itself. Spark made it unusually easy to move from natural-language instructions to a deployed full-stack experience, but GitHub now wants that work to continue closer to its broader Copilot development environment.


Anyone with an existing Spark project should treat the repository export as the priority. A live deployment continuing to work is useful, but having the source in a normal GitHub repository gives designers and developers much more control over future maintenance, hosting decisions, dependencies, and AI integrations.


We also see the llm() change as an important reminder about managed AI features. A generated web app can depend on platform services that are less visible than its interface, so exporting the code is only the first step; creators should also review external services and replace dependencies that will no longer be provided by GitHub.



Sources and Recommended Links