Natural Language Web (NLWeb) brings natural language interfaces to websites. Site content is ingested (e.g. Schema.org, RSS), turned into embeddings, and stored in a vector DB. Users ask in plain language ("Family hotel in Honolulu with a pool"); the NLWeb app interprets the query, retrieves relevant content, and answers from the catalog. The same setup can act as an MCP server, so external AI agents can call an "ask" tool to query the site. That makes the site part of the agent ecosystem.
NLWeb: conversation flow on a website
If the site is also an MCP server, an external AI agent can call an "ask" tool with the same query and get the same style of answer.
NLWeb: natural language on the web
Users ask in plain language ("Find me a family hotel in Honolulu with a pool"). The NLWeb app understands the query, searches the vector store for relevant catalog data, and returns an answer grounded in the site’s content. Because it also acts as an MCP server, external AI agents can call an "ask" tool to query the same site.
Example: Travel site with NLWeb
Catalog (flights, hotels, tours) is embedded and stored. User types: "Find me a family-friendly hotel in Honolulu with a pool for next week." NLWeb searches the vector DB, uses the LLM to interpret and format the answer, and returns real options from the catalog — no hallucinated listings. An external travel agent can also ask the site via MCP.