SaaS · archived
GitHub
GitHub is deprecating GitHub Spark, its prompt-to-app builder on github.com. As of 4 August 2026 it no longer accepts new users or new app creation; existing users have until 31 August 2026 to open Spark and export their app code before access ends.
left 31 Aug 2026
Quick answers
- Is it true that GitHub Spark is being retired — export your apps by 31 August 2026?
- Yes — and the date has already passed (31 Aug 2026). GitHub Spark was GitHub's "describe it and it builds it" tool: you wrote a prompt and Spark generated, hosted, and let you share a small web app, including AI features through a built-in llm() function.
- When did it take effect?
- 31 Aug 2026 (already passed). Verified against the vendor's own announcement.
- What should I use instead?
- If you built anything in GitHub Spark you want to keep working on, open it and export the app code before 31 August 2026 — that is the hard deadline, and once it passes you cannot get back into Spark to retrieve it.
What this means for you
GitHub Spark was GitHub's "describe it and it builds it" tool: you wrote a prompt and Spark generated, hosted, and let you share a small web app, including AI features through a built-in llm() function. GitHub is now winding it down on github.com. Two dates matter. Since 4 August 2026 Spark has stopped accepting new users and no longer lets anyone create new apps. Then on 31 August 2026, access to Spark itself ends — after that date you can no longer open Spark to edit or export the apps you built there. GitHub says apps you have already deployed will keep running after the retirement, so a live Spark app does not go dark on 31 August; what you lose is the ability to go back into Spark and change or retrieve the source. Separately, GitHub Models — the inference service that powered Spark's llm() function — was retired on 30 July 2026, so any Spark app that calls llm() has already stopped getting AI responses through it. GitHub frames the shutdown as a consolidation: it says the app-building experience Spark offered is now better served by GitHub Copilot in the tools developers already use (VS Code, the Copilot CLI, and the GitHub Copilot app).
Where to go
If you built anything in GitHub Spark you want to keep working on, open it and export the app code before 31 August 2026 — that is the hard deadline, and once it passes you cannot get back into Spark to retrieve it. Deployed apps continue to run, so there is no rush to redeploy, but the editable source lives inside Spark until you export it. If your app used the built-in llm() function, that path is already dead as of the 30 July 2026 GitHub Models retirement: to restore AI features you will need to wire the app to your own inference provider and manage your own API key and billing. For building new small apps from a prompt, GitHub points to Copilot — the agent modes in VS Code, the Copilot CLI, and the GitHub Copilot app — as the replacement workflow.
- GitHub Copilot in VS Code, the Copilot CLI, or the GitHub Copilot app — GitHub's named replacement for prompt-driven app building
- Any external LLM API (bring-your-own key and billing) to replace the retired llm() / GitHub Models inference that Spark apps relied on
GitHub Spark launched as one of the clearest expressions of the “vibe coding” idea inside GitHub itself: type a description of the app you wanted, and Spark generated it, hosted it, and gave you a shareable link — AI features included, through a built-in llm() function that needed no key of your own. Just over a year later, GitHub is retiring it on github.com.
The two dates that matter
The wind-down has already started and moves quickly:
- 4 August 2026 — Spark stopped accepting new users and no longer allows anyone to create new apps.
- 31 August 2026 — access to Spark ends. After this date you can no longer open Spark to edit your apps or export their code.
The important nuance is what survives. GitHub says apps you have already deployed will keep working after the retirement — a live Spark app does not switch off on 31 August. What disappears is the workshop: the ability to go back into Spark, change an app, or pull out its source. So the deadline is specifically an export deadline. If there is a Spark app whose code you might ever want again, the window to retrieve it closes on 31 August 2026.
The llm() function is already gone
There is a second, quieter deadline that has already passed. Spark’s built-in AI ran on GitHub Models, GitHub’s inference service, and GitHub Models was retired on 30 July 2026. Any Spark app that called llm() stopped getting responses through it on that date. Restoring AI features means pointing the app at your own inference provider and bringing your own API key and billing — there is no first-party replacement for the free built-in call.
Why GitHub says it is doing this
GitHub frames the retirement as consolidation rather than retreat. Its reasoning: since Spark launched, AI models and agentic coding tools have advanced enough that the same “idea to working app” loop is now better served by GitHub Copilot inside the tools developers already use — VS Code, the Copilot CLI, and the GitHub Copilot app — and it says that is where builders have been moving anyway. Whatever the merits, the practical takeaway is unchanged: if you have work inside Spark, get it out before 31 August 2026.
Entry changelog
- — last reviewed.
- — Entry created from GitHub's official changelog (4 August 2026). New users/app creation stopped 4 August; export access ends 31 August 2026; deployed apps keep running. Related: GitHub Models (the llm() backend) retired 30 July 2026. Surfaced via Layer-2 discovery.
- — The 31 August 2026 export-access cutoff has arrived; status moved to passed. Deployed Spark apps keep running, but the export path GitHub provided is now closed.
Put this countdown in your README
Building on GitHub? Drop this badge into a project’s README or docs — it shows the days left and
refreshes on its own every night from this page. 
[](https://lastcall.dev/entries/github-spark-shutdown)Spot an error or a changed date? Tell us and it’s corrected the same night.