Questions

Answers before the sales call

Seven questions every buyer asks about WebMCP, and the eight rules we work under. Several of the answers are no, not yet, or never guaranteed, which is exactly why they are published here rather than delivered in a call.

Send Me the Sample Report

We will send the sample report, then a short series on what an agent-ready site does and where agent journeys break. One click unsubscribes and we will not email again.

FAQ

The seven questions we are always asked.

Short answers. No hedging, and no pretending the technology is further along than it is.

Is WebMCP the same as MCP?

No. Backend MCP makes persistent data and operations available outside the website, on your servers. WebMCP exposes frontend tools tied to an open browser tab, the current page and its session. They are designed to complement each other, and they do not share a wire format.

Does WebMCP replace our APIs?

No. A WebMCP tool can call your existing APIs and application logic, but it does not replace the systems behind them. It is a declaration layer, not an architecture change.

Can every AI agent use WebMCP?

Not yet. Browser and agent support is still developing: it is in origin trial in Chrome and Edge, with no signal yet from Firefox or Safari. We implement it as progressive enhancement and retain ordinary HTML and UI interaction as the fallback, so unsupported agents and every human visitor are unaffected.

Will WebMCP improve our Google rankings?

Google has not identified WebMCP as a ranking signal, and we will not imply otherwise. Semantic improvements may help machines interpret a page more accurately, and that is worth doing on its own terms. Rankings are never guaranteed.

Do you need to redesign our website?

Usually not. The Retrofit is designed to enhance the existing site's semantics, controls and application logic. If we do recommend a design change, it will be because a specific journey cannot be completed without one, and we will show you the failing transcript.

Can agents purchase or submit forms automatically?

Read-only tools can often run automatically. Purchases, messages, bookings, applications and private-data submissions should require visible user confirmation. The browser does not enforce that boundary, so we build it into the page and we test that it holds.

What does ongoing monitoring test?

Whether agents can still understand the page, select the correct tools, complete important journeys, and stop at the required confirmation points. It also watches for structured-data regressions, claim drift and changes in browser support.

Trust and standards

Machine-first does not mean machine-only.

The constraints we accept before a client asks us to. If we ever break one, you are entitled to point at this page.

How we verify

The Semantic Surface rules

  1. Everything a machine is asked to believe must be available for a person to inspect.
  2. Everything an agent can do must be visible to the user.
  3. Consequential actions require confirmation.
  4. Structured claims must match the page.
  5. Tools must respect authentication and authorization.
  6. The ordinary website must keep working without agent support.
  7. Experimental APIs must be identified honestly.
  8. Ranking and conversion improvements are measured, never guaranteed.

Rule eight is the one that costs us work. We keep it anyway.

Can an agent finish the job on your website?

We will test the journeys that create revenue, show you where agents get stuck, and build the semantic and action layers needed to move them forward.