When I watch an AI agent or a coding tool interact with a website today, the process feels inefficient. These models are essentially forced to squint at raw HTML and minified JavaScript bundles while attempting to reverse-engineer the original intent of the developer. They miss the relationship between components, they lose the nuances of brand colors, and they struggle to understand the shape of underlying APIs. I built WebDNA to solve this by creating a bridge between the way humans write code and the way machines need to consume it.
The core idea is a manifest that lives at a predictable location on your server. By integrating directly into the Next.js build process, WebDNA automatically scans your project and generates a blueprint of your site. It looks at your route tree, extracts design tokens from your Tailwind configuration, and maps out your component hierarchy. This happens without requiring any manual documentation from the developer. It turns the implicit knowledge stored in your codebase into an explicit, structured briefing for any AI that visits your site.
This system moves beyond existing solutions like sitemaps or basic text files. While those tools offer a list of URLs or high-level summaries, they do not explain the design system or the specific data requirements of a dynamic route. WebDNA provides that missing context. It allows a developer to define which parts of a site are private and which components serve specific roles, ensuring that when an AI tool accesses the site, it does so with a full understanding of the constraints and the brand guidelines.
Security and privacy are foundational to this standard. The manifest is a static JSON file that cannot execute code and is read-only by design. I have included features like element-level exclusion, where a simple attribute can hide specific sections of a page from the AI manifest. This ensures that developers remain in total control of what information is shared. It is about providing consent-based, structured access rather than leaving AI tools to scrape whatever they can find.
My goal with WebDNA is to make the web more accessible to the next generation of software tools. By making the architecture of a site discoverable, we reduce the friction and errors that occur when AI tries to navigate or modify web projects. It is a simple addition to a configuration file that provides a significant upgrade in how machines perceive and interact with the work we build. Information about the package and the specification is available on GitHub and through the official documentation.
Official documentation: webdna.omraj.guru
Top comments (0)