Standard · Discover
NLWeb
A site answers natural-language questions from its own structured data.
Watching
What it is
An open project that lets a site answer plain-language questions at an /ask address, using the schema.org data the site already publishes. It can also expose that endpoint over MCP.
An analogy
A receptionist built from your own paperwork.
At a glance
- Owner
- Microsoft (MIT license)
- Maturity
- Watching Announced May 2025. We could not confirm a 2026 release or adoption by the large assistants.
- Where it lives
- NLWeb on GitHub
- Framework domain
- Discover: Be found
- What it serves
- Answer questions from your published data
Who should care
The roles this standard touches most. See all the roles on the personas page.
- Engineering and product. Build the access, data, endpoints and actions that agents use, on open standards.
How to adopt it
- 1
Publish good schema.org data first, since NLWeb depends on it.
- 2
Pilot it on one section if you want to learn.
- 3
Check whether the assistants you care about call it before investing further.
The Institute’s view
Watch
A sound idea without confirmed uptake. An MCP endpoint reaches more agents today. Revisit if a major search or assistant product starts calling /ask.
Standards change fast. Last reviewed November 2026.