WebMCPify Search · Launching soon

Search what the web can do.

Find WebMCP-enabled sites by the actions they expose, inspect the requirements, then visit the site or invoke the chosen capability through a supported agent path.

CAPABILITY SEARCH

Pages describe information.
Capabilities describe action.

Open Search preview ↗Choose an agent connection Submit a WebMCP site ↓

LAUNCHING SOON The interface works, but this deployment has no configured capability index or invocation backend yet.

SEARCH INDEX

Start with an intent.

Real capability records will appear here with source attribution, schemas, risk, authentication, freshness and verification signals.

No sample results are shown as real data.

The complete Search loop

Intent in. Verified capability out.

Search is designed to carry source and health information through discovery, selection and invocation—without turning an execution claim into proof.

01
Discover WebMCP-enabled websites.
02
Index sites, tools, schemas and descriptions.
03
Search by natural-language intent.
04
Rank relevant capabilities.
05
Inspect authentication, input and risk requirements.
06
Invoke a selected tool when supported.
07
Verify the outcome.
08
Record success, failure, latency and freshness.

WebMCPify Scout · In development

An independent crawler for the actionable web.

Scout is the planned discovery pipeline inside Search: find permitted public WebMCP surfaces, recheck them, and preserve where every record came from.

Independent crawl

Discovers public capability signals while respecting robots rules, origin boundaries and crawl limits.

Direct submission

Lets publishers request discovery while keeping submitter intent and first-party attribution visible.

Publisher feeds

Imports approved first-party records with provenance and freshness attached.

Ecosystem submissions

Lets maintainers contribute verifiable site and tool records under their applicable terms.

Browser verification

Revisits supported sites to distinguish a directory record from a currently observed runtime tool.

No unsupported brands

Live examples come from confirmed WebMCP records—not familiar sites that have no known integration.

Every indexed record keeps its source and freshness visible. Directory coverage is never represented as a live guarantee without a separate verification signal.

One search layer · four ways in

Meet agents where they already work.

Search is being designed as infrastructure, not just a destination page. Each access path keeps source, risk and availability visible.

Web preview

People can try the interface, inspect its status, and later open a matched website.

Native WebMCP

A compatible browser agent can discover the page-scoped Search tools; backend-dependent calls report unavailable honestly.

webmcpify-search

A planned MCP server for IDEs, coding agents, voice agents, and developer applications.

PLANNED

WebMCPify Guard

The open-source, user-owned security layer that protects the full action path and can connect MCP-compatible agents to website WebMCP tools.