isWebMCP
Daily adoption deskSeptember 10, 2026

WebMCP adoption report

Who is shipping tools—and what those tools actually do.

The first isWebMCP adoption baseline separates organizations from deployments, platform-inherited tools from provider engineering, and current document.modelContext implementations from legacy surfaces.

JSON MarkdownRSS

9

Provider-engineered

External census

72

Platform-inherited

One ReadMe bundle

4

Directly inspected

This report

28

Named tools

Evidence-backed records

Today's readout

The number needs a denominator

No material implementation change identified; all scheduled source rechecks reproduced their expected signals.

01

Deployment count overstates independent adoption

The best public census located for this report found 81 WebMCP-bearing provider sites, but attributed 72 of them to one ReadMe-hosted search_docs bundle. The more decision-useful number is therefore nine provider-engineered deployments, not 81 independent implementation decisions.

02

Discovery and documentation dominate the named tool set

The directly inspectable external surfaces concentrate on search, markdown retrieval, llms.txt retrieval, and catalog discovery. This is a low-risk entry point: read-only tools make documentation easier for an agent to navigate without authorizing consequential business actions.

03

Draft churn is already visible in production

The external census reported six deployments on the older navigator surface. Render's inspected homepage still selected navigator.modelContext on this report date. A registration can therefore exist in source while remaining unavailable to a client that implements only the current document.modelContext surface.

04

Runtime success remains a separate claim

Source inspection can confirm a registration path, tool names, and schemas. It does not prove that an enabled browser exposed the tools, that an agent selected them, or that invocation completed. This report does not upgrade source evidence into runtime verification.

Counting model

81 sightings ≠ 81 adoption decisions

The linked external census started with 730 MCP providers. Its 81 WebMCP detections split into two very different groups.

72

Inherited from ReadMe

One platform bundle, repeated across hosted docs.

9

Provider-engineered

The more meaningful early-adoption indicator.

Counts are reproduced from the linked APIs.io census and are not an independently replicated isWebMCP crawl. Daily reports retain that provenance until a first-party census is complete.

Implementation inventory

Who implemented what

Filter the evidence ledger by organization, tool, verification level, or API generation. Unknown means unverified—not absent.

7 of 7 organizations shown

Source confirmedCurrent APIprovider engineered

Cloudflare

Edge-injected WebMCP bridge plus page-owned tool manifests

Cloudflare announced a developer preview that injects a same-origin bridge at the edge and registers tool packs. Its homepage exposed a WebMCP JSON manifest with two read-only discovery tools when inspected for this report.

2

named tools

2

reported properties

Tools, sources, and limits

Tool inventory

  • read_llms_txt

    Read concise site context for assistants.

  • read_markdown_page

    Read the markdown representation of a Cloudflare page by path.

Evidence trail

Limits

  • The page manifest was inspected, but an enabled browser-agent invocation was not run.
  • The two-property count is inherited from the linked census.
Source confirmedLegacy APIprovider engineered

Render

Inline documentation tool registry

Render's homepage embedded five read-only documentation and content-index tools. The bootstrap selected navigator.modelContext and included a provideContext fallback, so the source is clear but tied to an older proposal surface on this report date.

5

named tools

1

reported property

Tools, sources, and limits

Tool inventory

  • render.docs.search

    Search Render documentation by keyword.

  • render.docs.get-markdown

    Fetch a Render documentation page as markdown by slug.

  • render.llms.get-index

    Fetch Render's llms.txt index as markdown.

  • render.blog.get-index

    Fetch the recent Render blog index as markdown.

  • render.articles.get-index

    Fetch the recent Render articles index as markdown.

Evidence trail

Limits

  • No enabled-browser discovery or invocation was run.
  • Legacy source presence may not produce tools in current-only clients.
Source confirmedCurrent APIprovider engineered

APIs.io

Page-local registry with current-surface preference and legacy fallback

APIs.io prefers document.modelContext.registerTool and falls back to the older navigator context. Its homepage registers catalog search and network-description tools and records capability and invocation telemetry.

2

named tools

2

reported properties

Tools, sources, and limits

Tool inventory

  • search_apis

    Search the federated API catalog by keyword, tag, provider, or artifact dimension.

  • describe_network

    Describe the APIs.io discovery network and machine-readable entry points.

Evidence trail

Limits

  • The implementation publishes its own census; figures are not treated as an independent audit.
  • The Stack page was not separately runtime-tested for this report.
Third-party observedMixed APIplatform inherited

ReadMe

Shared hosted documentation bundle

One ReadMe docs-hub bundle reportedly registered the same search_docs tool across 72 provider documentation sites. The platform should receive the engineering attribution; the hosted providers should not automatically be counted as 72 independent adoption decisions.

1

named tools

72

reported properties

Tools, sources, and limits

Tool inventory

  • search_docs

    Search the hosted provider's documentation.

Evidence trail

Limits

  • isWebMCP has not yet reproduced the complete 72-site inventory.
  • Hosted-provider awareness or intent cannot be inferred from platform injection.
Third-party observedUnknown APIprovider engineered

Zapier

Provider-engineered registration detected by an external source census

Zapier was named among provider-engineered deployments. Its public homepage included a WebMCP origin-trial token on this report date, but this report did not recover a complete first-party tool inventory from the loaded bundles.

—

named tools

1

reported property

Tools, sources, and limits

Tool inventory

Exact tool names remain unverified.

Evidence trail

Limits

  • Origin-trial enrollment alone is not proof of registered tools.
  • Exact tools and active API surface remain unverified by isWebMCP.
Third-party observedUnknown APIprovider engineered

Nylas

Provider-engineered registration detected by an external source census

Nylas was named among provider-engineered deployments. Its homepage loaded a same-origin MCP connector asset on this report date, but the asset name is not treated as proof of a current WebMCP tool surface.

—

named tools

1

reported property

Tools, sources, and limits

Tool inventory

Exact tool names remain unverified.

Evidence trail

Limits

  • Exact tools and active API surface remain unverified by isWebMCP.
  • A server-side MCP or connector label is not automatically WebMCP evidence.
Source confirmedCurrent APIprovider engineered

isWebMCP

Route-aware imperative tool registry using document.modelContext

isWebMCP registers a global evidence, learning, news, demo, and export surface, plus three route-scoped product-demo tools. This record is self-reported from the same repository that publishes the adoption report and should not be treated as independent validation.

18

named tools

1

reported property

Tools, sources, and limits

Tool inventory

  • scan_public_url

    Scan bounded public source evidence.

  • get_scan_summary

    Read the active scan summary.

  • list_action_surface

    List candidate actions from a scan.

  • get_finding_details

    Read evidence for one finding.

  • search_webmcp_library

    Search reviewed WebMCP resources.

  • get_webmcp_resource

    Read a reviewed WebMCP resource.

  • get_webmcp_adoption_report

    Read the latest or a dated WebMCP adoption report.

  • list_webmcp_adopters

    Search the latest adoption ledger by organization or tool.

  • list_mcp_updates

    List source-linked ecosystem changes.

  • get_challenge_pulse

    Read archived event observations.

  • select_demo_mode

    Select and reset a synthetic demo mode.

  • start_demo_run

    Start a controlled synthetic journey.

  • finish_demo_run

    Finish and score a synthetic journey.

  • compare_demo_runs

    Compare paired synthetic runs.

  • export_current_report

    Prepare a JSON or print report export.

  • search_demo_products

    Search the synthetic product catalog.

  • compare_demo_products

    Compare selected synthetic products.

  • add_demo_product_to_cart

    Update the synthetic visible cart.

Evidence trail

Limits

  • This is a self-report from the publisher of this adoption report.
  • Source confirmation is not an independent runtime trial.

Evidence policy

Claims stop where observation stops

Public WebMCP browser-tool adoption. Remote MCP servers are excluded unless a page bridges them into document.modelContext or the legacy navigator surface.

Read the wider evidence methodology

runtime verified

An enabled browser discovered and invoked the tool with a recorded result.

source confirmed

First-party page or repository source contains the registration path and enough definition material to identify the implementation.

third party observed

A linked external census reports source detection that is not yet independently reproduced by isWebMCP.

announced

A first-party announcement describes support, without source or runtime confirmation in this report.

Append-only archive

Daily reports

1 published snapshot

No material implementation change identified; all scheduled source rechecks reproduced their expected signals.