Microsoft has launched a new open-source project called NLWeb, which claims to transform any existing website into an AI-powered application by integrating natural language capabilities.
The project was announced at Microsoft Build 2025, aiming to simplify the creation of natural language interfaces that can be embedded into any website.
NLWeb, or Natural Language Web, is designed to provide developers with the quickest and easiest way to turn any website into an AI application powered by their chosen large language model. Once integrated, users will be able to query any website's content via voice, similar to using AI assistants like ChatGPT or Microsoft Copilot.
The company states that each NLWeb instance serves as a model context protocol server, making a website's content discoverable and accessible by AI agents. Ultimately, it envisions NLWeb becoming for agent AI what the HTML protocol is for accessing the internet through web browsers.
Microsoft explains that NLWeb utilizes a website's already-published semi-structured data, such as Schema.org and RSS feeds. It combines these with large language models to create a natural language interface accessible to both humans and AI agents.
Semi-structured data will be enhanced by external knowledge from the underlying large language models that support the natural language interface. For example, when someone queries about restaurants, it will be able to overlay geographic insights, the company says.
As the project is entirely open-source, it is technology-agnostic, meaning it supports major operating systems beyond Windows, such as Android, iOS, and Linux. Developers can choose components that best suit their needs, including the underlying large language model and vector database to enhance search experiences.
Microsoft says NLWeb aims to bring the benefits of generative AI search directly to every website. Just as HTML enabled almost anyone to create and access websites using a browser, NLWeb aims to make it easy for publishers to create intelligent natural language experiences to help users navigate their sites better. It is also being built with future AI agents in mind, with Microsoft believing these agents will soon account for a significant portion of internet traffic.
NLWeb was conceived and developed by Microsoft Technical Fellow R.V. Guha, the creator of widely-used web standards like RSS, RDF, and Schema.org. The project already has numerous contributors involved in its development.
The company emphasizes making NLWeb as easy as possible to implement with any website. It is based on a lightweight codebase controlling the core service that processes natural language queries and provides documentation on how to extend and customize it.
In addition, there are dozens of connectors for popular AI models and vector databases, along with documentation explaining how to add models not on the official list. There is also a set of tools to directly add Schema.org, RSS, and other format data to any vector database, as well as web server frontends and user interfaces that can be directly integrated into any website.
Many websites have already deployed NLWeb, including Tripadvisor, Shopify, and Eventbrite. Contributors include organizations like Chicago Public Media, Common Sense Media, Inception Labs, O’Reilly Media, and Snowflake Inc.
