GitHub launches 'Spark' to build full-stack apps from plain text
GitHub has introduced Spark, an AI tool that builds full-stack applications using plain-language prompts, now in preview for Copilot Pro+ users.
GitHub has unveiled Spark, an AI-powered development tool that allows users to generate and deploy full-stack web applications from plain-language prompts.
Currently available in public preview for Copilot Pro+ subscribers, Spark aims to streamline app creation for developers and non-technical users alike by using natural language inputs in place of traditional code.
What does Spark enable?
Spark introduces what GitHub refers to as “vibe coding”, which is a conversational development approach where users describe an app’s functionality, interface, and features in English.
Spark then translates these prompts into functioning code, creating front-end and back-end logic, database connections, and deployment infrastructure automatically.
According to GitHub, once a prompt is submitted, Spark generates a full GitHub repository containing all the necessary files, which can be edited, refined, and deployed with one click.
This process includes integration with GitHub Actions for continuous deployment and support for custom APIs, authentication flows, and AI-based features.
Spark supports iteration, meaning users can continue to modify the app’s design or logic by updating their prompts or editing the code within the generated repository. GitHub emphasises that Spark maintains version history and commit structures, aligning with conventional developer workflows.
How is Spark unique?
While other AI-powered development platforms, such as Replit’s Ghostwriter, also offer code generation from text, GitHub Spark’s tight integration with the GitHub ecosystem sets it apart.
Every Spark project is natively built and managed on GitHub, enabling collaboration, branching, and pull request workflows.
In a statement, GitHub described Spark as “prompt to production,” highlighting the tool’s ability to create fully deployable applications, not just code snippets or prototypes.
Spark is believed to leverage the capabilities of Claude Sonnet 4 or equivalent large language models, though GitHub has not officially confirmed the specific AI model behind the service.
The tool supports both web and mobile application development and is currently optimised for JavaScript-based stacks.
GitHub has not yet announced a general release date for Spark beyond the Copilot Pro+ preview, but further updates are expected later this year.


