Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

Microsoft’s NLWeb project aims to let website visitors ask questions about a publisher’s content in natural language. Announced on May 19, 2025, NLWeb is an open project that combines site data with a language model selected by the developer; publishers can also choose whether to expose their content to AI agents through the Model Context Protocol (MCP). It is a project with an ambition, not an established web standard or a proven measure of adoption.

What is NLWeb?

NLWeb, short for Natural Language Web, is Microsoft’s open project for adding a conversational interface to a website. Instead of navigating pages or searching with keywords alone, a visitor could ask a question in ordinary language and receive a response based on the site’s content. Microsoft announced the project at Build on May 19, 2025. Microsoft’s announcement describes the goal as making a natural-language experience available to web publishers.

Microsoft compares the ambition to HTML’s role in making websites easier to create. The comparison expresses the company’s goal; it does not establish that NLWeb has become a standard or reached HTML-like adoption. Microsoft Corporate Blogs put the goal this way: “Just like the introduction of HTML made it easy for almost anyone to create a website, we want NLWeb to make it easy for any web publisher to create an intelligent, natural language experience for their site.”

How NLWeb is intended to work

It uses a site’s existing content

Microsoft says NLWeb can draw on semi-structured data that publishers may already provide, including Schema.org markup and RSS feeds, along with other website data. LLM-powered tools use that material to support natural-language queries. The usefulness of the result therefore depends in part on what data a site makes available and how suitable it is for retrieval; the announcement does not establish uniform results across websites.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

Developers choose the components

Microsoft describes NLWeb as technology-agnostic, with support for major operating systems, models, and vector databases. Developers can select the model and retrieval infrastructure rather than relying on a single prescribed combination. Microsoft’s overview describes this design, but does not independently compare the performance of particular models or databases. The NLWeb project overview outlines the project’s technical approach.

The visitor interface and agent access are different capabilities

Microsoft says each NLWeb instance is also an MCP server. That can make a publisher’s content accessible to AI agents, if the publisher chooses to enable that use. It is separate from the visitor-facing natural-language interface: an NLWeb-powered experience does not mean every website automatically grants agents access to its content. Microsoft’s description of the NLWeb stack presents MCP as part of the project’s design.

What developers get, and what they still need to decide

Microsoft’s getting-started description says the project repository includes core service code and extension documentation, connectors for models and vector databases, tools for loading data, and a web-server frontend with a basic query interface. It lists Schema.org, JSONL, RSS, and other formats as supported data inputs. That is a starting point for implementation, not a guarantee that setup will be effortless for every site. The repository’s getting-started material describes these components.

Before implementing NLWeb, a publisher or developer needs to make practical choices:

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.
  • Site data: Identify what structured or semi-structured content is available and whether it is current and useful for answering visitor questions.
  • Model: Choose the language model to use with the site’s content.
  • Retrieval backend: Select a vector database or other retrieval components suited to the implementation.
  • Agent access: Decide whether to expose content to agents through MCP, separately from offering a natural-language interface to visitors.

Microsoft establishes that these choices are available, but its announcement does not identify one best configuration or provide independent performance comparisons.

Who was involved at launch?

In its May 19, 2025 announcement, Microsoft named these launch-era collaborators: Chicago Public Media; Common Sense Media; DDM (Allrecipes/Serious Eats); Eventbrite; Hearst (Delish); Inception Labs; Milvus; O’Reilly Media; Qdrant; Shopify; Snowflake; and Tripadvisor. Microsoft described a “small cohort” of early adopters but did not give a count. The announcement does not establish that every named organization remains involved or runs NLWeb in production today. The launch announcement lists the collaborators.

Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Support on Ko-Fi

Why is Microsoft pursuing NLWeb?

Microsoft’s stated product aim is to make natural-language interaction with website content easier for publishers, while its MCP design can also make that content available to agents when publishers opt in. Those are the capabilities Microsoft describes; the announcement does not establish commercial outcomes or measured adoption.

A contemporaneous Computerworld analysis published May 20, 2025 discussed Microsoft’s interest in the agentic web. That is an outside strategic interpretation, distinct from Microsoft’s own description of NLWeb’s purpose.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.