Google has introduced a new protocol called WebMCP, aimed at helping websites better support AI agents that are increasingly performing tasks on behalf of users, from booking flights to completing online purchases.
Currently available in early preview, WebMCP is designed to provide a standardized way for websites to expose structured tools and actions to AI agents. The move signals a shift in how the open web is being adapted for agent-driven interactions.
A Google spokesperson said the protocol creates a direct connection that allows AI agents to execute complex tasks with greater speed, reliability, and precision. As AI systems move beyond answering queries to actively completing workflows, websites need clearer mechanisms to guide those interactions.
At present, most AI agents navigate websites much like human users: clicking through pages, scanning buttons, and attempting to fill out forms step by step. This process can be slow and error-prone, particularly for multi-step actions such as checkout flows or booking systems.
WebMCP addresses this by allowing websites to define a set of structured, predefined actions that agents can access directly. Instead of guessing which elements to interact with, agents receive explicit instructions about available tools and workflows.
“The goal is to create a standard way for exposing structured tools, ensuring AI agents can perform actions on your side with increased speed, reliability, and precision,” wrote André Cipriani Bandarra, a staff developer relations engineer at Google, in a company blog post.
By defining these tools, websites effectively tell AI agents how and where to interact, reducing ambiguity and improving performance. The protocol creates a more robust communication channel between websites and AI systems.
The potential applications span several industries. Ecommerce platforms could use WebMCP to streamline product searches and checkout processes for AI agents acting on behalf of customers. Travel providers could enable more reliable flight searches and booking flows. Customer support systems could allow agents to generate and submit detailed tickets automatically using structured inputs.
Google describes WebMCP as an open, complementary protocol designed to integrate smoothly with existing web stacks. Unlike approaches that require deeper backend server-side integration, WebMCP allows sites to implement agent capabilities using their current infrastructure.
The rollout reflects a broader transformation of the internet’s underlying architecture as AI agents become more autonomous and transactional. If agents are to search, transact, and execute tasks reliably on behalf of users, web infrastructure will need to evolve accordingly.
With WebMCP, Google is positioning itself at the center of that shift, offering what it sees as a foundational layer for an internet increasingly built for AI-driven action rather than just human browsing.
