A WebMCP implementation that is correct at launch does not automatically stay correct indefinitely. A theme update, a new caching module, a hosting platform change, or even a CDN migration can affect how AI agents discover and use the tools exposed by your site, without anyone touching a single line of WebMCP code. The WebMCP maintenance service offered by GOAI consists of periodic checks that catch these degradations before they turn into problems that are hard to trace.
Why a correct implementation can degrade without being touched
Most web projects do not stand still. A content management system receives security updates, themes change to match a new design, and infrastructure teams introduce new caching or firewall rules to reduce server load. Each of these changes can, indirectly, affect the WebMCP discovery files, the manifest, or the server's responses to requests from AI agents. The problem is not visible to the internal team, because no one intentionally touched the WebMCP implementation. It was affected collaterally, by a decision made in an entirely different part of the project.
What we monitor periodically as part of WebMCP maintenance
The checks we perform are not theoretical, they are based on testing in a real Chrome environment. Among the elements we monitor periodically are:
- The number and names of tools actually registered in Chrome, compared against the reference list established at implementation;
- The completeness of tool schemas: parameters, data types, required descriptions;
- The validity of the origin trial token and how close it is to its expiration date, to avoid an unannounced interruption;
- How the manifest and discovery files are served: HTTP response code, content type, presence at the expected location;
- New blockages appearing in robots.txt, firewall rules, or WAF configurations that could prevent AI agents from accessing the resources they need.
These checks do not guarantee that a tool will be used by any particular AI agent, nor that your site will appear more often in a conversational assistant's responses. They only guarantee that the technical infrastructure remains functional and accessible, which is a necessary but not sufficient condition for any subsequent outcome.
How we notify you when something breaks, and why there is no visible symptom
A human visitor arriving on the site will notice nothing unusual if the WebMCP manifest stops loading or if a tool has lost part of its schema. The page looks the same, navigation works the same, forms work the same. The degradation is invisible from the perspective of ordinary browsing, because it affects only the channel through which AI agents interact with the site, not the visual channel meant for people. That is why the only reasonable way to detect such a problem is periodic, automated verification, not reports from users, who simply have no way of noticing it. When we identify a deviation from the reference state, we send a notification describing exactly what changed, since when we estimate the problem has existed, and how urgent we consider it to be, ranging from a simple heads-up to an issue requiring rapid intervention.
What the fix includes and what is discussed separately
Maintenance includes fixing corrective issues: repairing an incomplete schema, updating the manifest to the correct format, flagging the origin trial token and, where possible, renewing it, or adjusting a robots.txt or firewall rule that is accidentally blocking a legitimate AI agent. These interventions are part of the maintenance subscription and do not require a separate commercial discussion for each individual case.
What maintenance does not include, by its very nature, are structural changes: adding new tools, extending WebMCP to cover a product catalog introduced later, or redesigning the architecture used to expose tools. These requests are treated as separate projects, similar to a new WebMCP implementation service, with a dedicated discussion and quote, because they involve architectural decisions rather than simply correcting a deviation.
Change log
For every implementation under maintenance, we keep a change log recording the date of each check, the state found for every monitored element, and the actions taken where applicable. This log is made available to you and offers an overview of how the implementation has evolved over time, which is especially useful when your internal team changes the hosting platform or the site theme and wants to know whether that change had any effect on WebMCP.
Who this service is useful for
WebMCP maintenance makes the most sense for sites that have already gone through a full implementation or an AI agent readiness audit and want to keep that investment working over the long term, without relying on the internal team's memory of WebMCP's technical details. It is also useful for sites using an origin trial token with a known expiration date, where an early warning can prevent an unpleasant service interruption.
Request a quote for WebMCP maintenance
If you would like to discuss maintenance and monitoring for your WebMCP implementation, you can reach us through the contact page. Before deciding whether periodic maintenance is necessary, you can always check the current state of your site with the free WebMCP validator.