Skip to content
CitationLab
Back to Glossary
Glossary

What is NLWeb? Making your website agent-ready

NLWeb is an open project from Microsoft that makes websites conversational and queryable by AI agents — using schema.org and other existing web formats together with the Model Context Protocol. Every NLWeb instance can act as an MCP server. Here is what NLWeb is, how it relates to agent-ready websites, and what CitationLab does with it.

KR
Krister Ross
Founder & CEO, CitationLab
Published 2 min read
Curious how AI talks about your brand?Run a free visibility check

NLWeb is an open project from Microsoft, launched in May 2025, that makes it easy to add a conversational, AI-queryable interface on top of an existing website. Instead of an AI having to guess its way through navigation, footers and styling, NLWeb gives agents a direct, natural-language entry point to what the page is actually about. The goal is to turn your website into something an agent can query, not just something a human can read.

How NLWeb works

NLWeb reuses web formats you probably already have: schema.org, RSS, sitemaps and feeds. These are combined with an LLM and a vector database to produce grounded answers — answers anchored in the page's own content, not the model's free guessing. When a user or agent asks a question, NLWeb retrieves the relevant pieces of the site's structured data and phrases a natural answer sourced from your material.

NLWeb and MCP: the website as an MCP server

The key move is this: every NLWeb instance can act as an MCP server. It exposes the site's content to AI agents via endpoints such as /ask (for questions) and /mcp (for the Model Context Protocol). Your website becomes a service in the agentic ecosystem — an agent can connect and pull answers directly, the same standardized way it connects to any other MCP server.

Why it matters for AI visibility

Early adopters include Shopify, Snowflake, O'Reilly Media, Tripadvisor, Eventbrite and Hearst — a signal that the "agentic web" is becoming real infrastructure. For brands, NLWeb raises the bar: it is no longer enough for content to be pretty for humans, it has to be queryable by agents. Well-structured schema.org data then stops being an SEO add-on and becomes the very interface an agent uses to understand and recommend you.

How CitationLab works with agent-ready websites

Citation Sites — CitationLab's website product — are built agent-ready from the ground up: semantic HTML, rich schema.org structure and machine-readable content, so both classic AI crawlers and newer agent interfaces like NLWeb have something solid to pull. CitationLab also publishes llms.txt and agents.md for its own surfaces. Want to make your website agent-ready? Read more about AI visibility or look up the terms in the glossary.

Frequently asked questions

What is NLWeb?
NLWeb is an open project from Microsoft that makes it easy to add a conversational, AI-queryable interface to a website. It uses web formats you already have — schema.org, RSS, sitemaps and feeds — together with an LLM to answer questions grounded in the site's own content.
How does NLWeb relate to MCP?
Every NLWeb instance can act as an MCP server. That means your website can not only answer people in a chat, but also expose its content to AI agents via the Model Context Protocol, through endpoints such as /ask and /mcp. The site becomes both readable for people and queryable for agents.
What do I need to use NLWeb?
Well-structured content is the starting point: schema.org markup, sitemaps, RSS or feeds that describe your pages in a machine-readable way. NLWeb combines these with an LLM and a vector database to give grounded answers. The better your structured data, the better the answers.
Is NLWeb the same as an agent-ready website?
NLWeb is one way to make a website agent-ready, not the whole definition. An agent-ready website is a broader idea covering semantic HTML, schema.org, llms.txt, agents.md and clean APIs. NLWeb adds a concrete conversational and MCP layer on top, but the underlying principles of structured, machine-readable content are the same.
Key terms

Definitions used in this article

NLWeb
NLWeb is an open project from Microsoft that makes websites conversational and queryable by AI agents by combining existing web formats such as schema.org with an LLM, where each instance can also act as an MCP server.
Schema.org
Schema.org is a shared vocabulary of structured data that describes what content on a page actually is — product, article, organization, FAQ — so machines and AI can interpret it unambiguously.
Agent-ready website
An agent-ready website is a site built so AI agents can read, understand and act on its content — with semantic HTML, schema.org, llms.txt, agents.md and clean interfaces.
Share:

Hold deg oppdatert

Få fagartikler, produktnyheter og analyser rett i innboksen.