Mcp is an npm package published by ai.foura that gives AI agents a way to retrieve web pages reliably. It supports two retrieval modes: a proxy-based fetch for lightweight requests and a full browser rendering path for pages that rely on JavaScript or actively resist automated access with anti-bot measures. This makes it suited for agents that need to read content from sites that block or challenge standard HTTP clients.
The server communicates over stdio transport and requires a single environment variable, FOURA_API_KEY, to authenticate with the Foura service. Because the scraping infrastructure runs remotely, the agent does not need to manage headless browser dependencies locally; it delegates the fetch to the Foura backend.
No tools are enumerated in the registry record for this version (0.4.8), so the exact callable surface is not documented here. Based on the server description, the core capability is fetching arbitrary URLs, with the backend handling proxy rotation and browser execution on behalf of the agent.
This is the wrong choice if an agent only needs to call structured APIs that already provide JSON responses, or if the deployment environment cannot reach external services over the network. It is also not a general-purpose database or file-system tool.