Web MCP
Give your AI agent live search and readable web pages through one hosted MCP connection.
Web MCP connects your AI agent to WebScrapingAPI through the Model Context Protocol. Add a server URL and a dedicated MCP token to a compatible client. There is no package to install or server to run.
The agent can discover pages, read their content as Markdown, and retrieve more content from a saved page. Your agent chooses the sources and writes the final answer with citations.
Account access
To connect, your dashboard must show a configured server URL under MCP. MCP uses your existing services and quotas. Creating a token does not activate a product or add an allowance.
Connect once
The hosted endpoint is https://mcp.webscrapingapi.com/mcp. Use a client that supports Streamable HTTP with an Authorization: Bearer header.
Follow Getting started to create a token and connect your agent.
From a question to a cited answer
- Your agent calls
web_searchto find relevant URLs and snippets. - It selects sources and calls
web_readto retrieve readable page content. - It calls
web_read_morewhen it needs the next chunk of a large page. - It answers using the retrieved evidence and links to the supporting pages.
The MCP server supplies content and source metadata. Research decisions and answer generation happen in your agent. If you already have a URL, the agent can start with web_read without a search.
Tools and required services
| Tool | Purpose | Account access required |
|---|---|---|
web_search | Find Google search results with titles, URLs and snippets | SERP API |
web_read | Retrieve a public web page as Markdown | Web Scraping API or Browser API |
web_read_more | Continue reading the same saved page | Page-reading access and an available capture |
Tools are advertised according to the services available on your account. Without SERP access, you can still read known URLs through an enabled reading service. See Usage and errors for service availability and billing.
Automatic page routing
For ordinary reads, your agent only needs a URL. Web MCP uses Web Scraping API when available, otherwise Browser API. If you have both, Web Scraping API is the default. Advanced callers can request a specific available service.
Web Scraping API uses your credits and offers optional JavaScript rendering. Browser API always renders JavaScript and uses your uploaded and downloaded bandwidth. A failed read does not automatically switch services or consume another service's allowance.
Where to go next
- Getting started: token creation, connection settings and your first request.
- Tool reference: arguments, response fields, continuation and citation guidance.
- Usage and errors: account billing, limits and troubleshooting.