WebMCP is a specification that allows web developers to expose JavaScript functions and HTML form elements as tools with natural language descriptions and structured schemas for AI agent consumption.
WebMCP addresses the limitations of backend integrations for interactive web applications. When AI agents interact with services through traditional backend APIs, they bypass the web UI, lose user context, and require developers to replicate authentication and state on separate servers. WebMCP solves this by letting developers define tools directly in browser pages using JavaScript or HTML forms. Agents can then invoke these tools while maintaining access to the page's UI and shared context, enabling more reliable and direct interaction with web content originally designed for human users.
Developers should choose WebMCP when building web applications that benefit from AI agent interaction while preserving the user's visual experience and application state. It suits scenarios where agents need to actuate web content cooperatively with users rather than through separate backend servers. The project provides TypeScript type definitions through an npm package and documents implementation status across different browsers and agents, along with best practices for designing effective tools.
The project maintains active specification development with regular updates to the core documentation and implementation guidance. The team engages with browser and agent platform developers to track and improve compatibility across different environments. Development focuses on establishing WebMCP as a practical standard for in-browser agent tooling rather than on rapid feature expansion.
