PlateTaleWebMCP vs APIs vs browser agents: which should you support?
There are three routes an agent can take into your business, and they are not competing so much as stacking. The question is which one you invest in first.
Browser agents
The agent drives a real browser against your public site. It needs nothing from you, works today, and is how most agent traffic actually arrives. It is also the least reliable route, because it depends on an interface built for eyes.
APIs
The most reliable and the most work. An API gives an agent deterministic actions and real status codes, but it needs to be built, documented, versioned, authenticated and kept in sync with the product. Most consumer-facing businesses do not have one that covers their customer journeys.
WebMCP and action declaration
The middle position: the page itself advertises the actions available on it, so an agent invokes a declared tool instead of guessing which button to click. It reuses your existing site and auth rather than replacing them. It is early, and support is not universal.
How to choose
Practical order for almost everyone: make the browser path work first, because that is where the traffic already is. Declare actions for the handful of journeys that carry real value. Build or open an API where a partner needs volume and guarantees.
Whichever routes you support, the open question is the same — does the journey actually complete? That is measurable, and it is what PlateTale measures.