Nicolas Sitter
Live project

Agent tools — and who actually calls them

This site publishes its research and its AI-hotel-visibility data as tools an AI agent can call, over two surfaces. Every call is logged. The counts below are live.

Why

There is a great deal written about making a website “agent-ready” and almost no data on whether anything then turns up. Our own census of the MCP ecosystem found that 48% of probed travel and hotel MCP servers expose zero working actions — plenty of announcements, far fewer live endpoints. The obvious next question is what happens after you build one.

So this page is the experiment run on ourselves. The tools are real, the logging is unconditional, and the number is whatever it is. If the answer stays almost nobody for months, that is the finding, and it is more useful than another guide telling you to ship an MCP server.

Live data

Tool calls
1
Succeeded
100%
1 of 1
Distinct clients
1
Last call
11 Sept
since 11 Sept 2026

Which clients

ClientCallsTools
verification-probe11

An MCP client names itself in initialize. “unidentified” means it didn’t, which is normal for a hand-written request.

Which tools

ToolCallsAvg ms
initialize1

Most recent calls

ToolSurfaceClientRowsms
initializemcpverification-probe

How

MCP server

Any MCP client, anywhere. JSON-RPC 2.0 over HTTP, stateless, no auth.

https://nicolassitter.com/api/mcp
WebMCP

The same tools registered into the page via document.modelContext, for an agent driving a browser tab. Chrome-only and pre-standard.

Both surfaces read one registry, so a tool exists on both or on neither. Every tool goes through this site’s own public HTTP API rather than the database — they cannot return anything you could not fetch yourself.

The 8 tools

ToolReads
search_research_articlesSearch Nicolas Sitter's published research on AI search, answer engines and hotel technology by …
get_articleFetch one research article by its URL slug — the last path segment of a /research/<slug> URL…
get_ai_hotel_landscape_manifestRead the manifest for the AI Hotel Landscape dataset: which AI platforms are tracked (ChatGPT, G…
get_ai_hotel_visibilityGet the hotels an AI platform actually recommends for a destination in a given week, ranked by h…
get_hotel_brand_leaderboardGet the chain-brand leaderboard for one AI platform in one week: which hotel groups (Marriott, A…
get_ai_citation_sourcesGet what an AI platform cites when it answers hotel questions — the mix of source categories (OT…
get_chain_vs_independent_mixGet the share of chain hotels versus independents versus vacation rentals in what each AI platfo…
list_race_explorersList the endurance races this site publishes results explorers for — L'Étape du Tour, the Marath…

What is logged

Recorded
  • Tool name, surface, success, duration
  • The client’s self-reported name and version
  • Truncated user agent
  • Whitelisted parameters only, capped at 200 characters
Never recorded
  • IP addresses
  • Cookies, session ids, fingerprints
  • Any parameter not on a tool’s own whitelist
  • Anything at all about human readers of this page

The whitelist is per tool and additive by choice: a tool that gains a parameter does not start logging it until someone says so.