GitHub Sets August 31 Deadline for Spark App Exports as It Retires the AI App Builder

Overview

GitHub is retiring Spark, the natural-language app-building tool it built into github.com, according to a GitHub Changelog post published August 4, 2026. As of that date, “GitHub Spark no longer accepts new users or allows the creation of new apps,” and existing users have until August 31, 2026 “to export apps they’ve already created,” per the same post.

What We Know

Spark launched in public preview on July 23, 2025 for Copilot Pro+ subscribers, pitched as a way to go from “idea to deployed app in minutes,” according to GitHub’s launch changelog post. At launch, Spark let users “describe your idea and watch Spark build it, with frontend and backend capabilities included, all powered by Claude Sonnet 4,” and offered AI features “powered by LLMs from OpenAI, Meta, DeepSeek, xAI and more” with “no API key management needed,” the launch post said.

According to GitHub’s documentation on Spark, the tool let users “describe what you want in natural language and get a fullstack web app with data storage, AI features, and GitHub authentication built in,” then “deploy with a click to a fully managed runtime.” The docs describe Spark’s architecture as “an opinionated stack (React, TypeScript)” with a managed data store that “runs on Azure (Cosmos DB)” for records “up to 512 KB per entry,” and apps were “hosted and deployed by Azure Container Apps (ACA).” Access was limited to “Copilot Pro+ and Copilot Enterprise” subscribers.

GitHub’s stated reason for the shutdown is that “AI models and agentic development tools have advanced significantly” since Spark’s launch, and that “builders can now build and refine these experiences through GitHub Copilot in the environments where they already work, including VS Code, Copilot CLI, and the GitHub Copilot app,” the August 4 changelog post said. GitHub added that it has “seen builders increasingly choose these integrated workflows for application development,” and that the change “applies specifically to the current GitHub Spark experience on github.com.”

The deprecation compounds an earlier cutoff: GitHub Models, the inference service that powered Spark’s built-in llm() function, was fully shut down on July 30, 2026, as previously reported. According to GitHub’s retirement changelog, “the playground, model catalog, inference API, and bring your own key (BYOK) are no longer available to any customer, including existing customers with active usage,” and GitHub pointed users toward Microsoft Foundry or GitHub Copilot instead. The August 4 Spark post confirms the knock-on effect: “As of that date, calls to llm() no longer work. Apps that don’t use llm() aren’t affected by the GitHub Models retirement.”

GitHub laid out what current Spark users need to do. “Existing apps you’ve already deployed will continue to work after GitHub Spark shuts down,” the changelog said, but anyone who wants to keep editing an app must “export your app code before August 31, 2026” by opening the Spark workbench, selecting the menu, and choosing “Create repository.” For apps that call llm(), GitHub says to “replace it with your own inference provider to keep AI features working,” and that “users will need to supply their own API key and manage billing going forward in this instance, as GitHub will not provide the underlying model or tokens any longer.” GitHub also suggested checking whether an app is affected by searching its code for llm() calls: apps with none “aren’t affected” and need no changes.

What We Don’t Know

GitHub’s changelog post does not disclose how many apps were built on Spark, how many active users the platform had, or what happens to apps whose owners miss the August 31 export deadline beyond losing the ability to edit them in the Spark interface. The post also does not name a successor “Spark”-branded product, only pointing generally to Copilot’s existing surfaces in VS Code, Copilot CLI, and the GitHub Copilot app.