GitHub Spark, an artificial intelligence (AI) tool available within the GitHub Copilot, was released on Thursday. It is a vibe coding tool that allows users to generate apps via natural language descriptions. The Microsoft-owned platform also allows users to collaborate with the AI to better control the outcome in multiple ways.
Powered by Anthropic’s Claude Sonnet 4, the tool can generate both the backend and frontend capabilities of the app, and it can be directly published once the developer is satisfied with the end result.
GitHub Spark Is Now Available to Copilot Pro+ Users
In a post on X (formerly known as Twitter), Microsoft CEO Satya Nadella announced the release of GitHub Spark. The AI-powered tool was first announced in October 2024, and it is now available in public preview to GitHub’s Copilot Pro+ subscribers.
With this tool, available within Copilot, developers can generate micro apps dubbed Sparks. These micro apps are fully functional and come with both backend and frontend capabilities. The app generation is powered by Claude Sonnet 4, according to GitHub changelog.
Additionally, developers do not require setup for this, with the data, large language model (LLM) inference, hosting, deployment, and GitHub authentication all available right within the tool.
Developers can also add AI features into these apps. These features can be powered by third-party AI models from OpenAI, Meta, DeepSeek, xAI, and more. Users can directly integrate these models’ capabilities without needing an application programming interface (API) key.
GitHub Spark also allows developers to collaborate with the AI to build their desired apps. If they feel that the first iteration is not satisfactory, they can add additional natural language commends or directly change the frontend elements via visual controls. Alternatively, developers can also use GitHub Copilot’s code completions to write, edit, or change code snippets.
Developers can also create a repository with GitHub Actions and Dependabot with just a single click. This way, the entire project remains synchronized. Additionally, by opening codespace, they can ask Copilot agent to iterate on the Spark and build on top of it.
