MCP integration guide / 2026
Hosted MCP Server for Web Search and Extraction in 2026
MCP clients can discover and call web-data tools without a custom SDK. A hosted server is useful when an agent needs both known-URL extraction and focused discovery while retaining source-linked results.
Quick answer
Connect agents to one hosted MCP endpoint.
Configure the hosted MCP URL to expose public-page extraction plus web, news, image, and place search as agent tools. The server is stateless and follows the same cache, allowance, and credit behavior as the public API.
{
"mcpServers": {
"extractor": { "url": "https://extractor.sh/mcp" }
}
}Available data
What you can extract
- Public URL extraction tools
- Web and news search tools
- Image and place search tools
- Markdown or normalized structured results
AI workflows
Where normalized data helps
- Coding and research agents
- Desktop AI assistants
- MCP-enabled RAG workflows
- Source-grounded tool use
AI-ready output
Markdown for models. JSON for systems.
Raw HTML consumes tokens on navigation, scripts, styling, and interface labels. Clean Markdown keeps the readable hierarchy for LLM prompts and RAG chunks. Normalized JSON is better when your application needs an explicit semantic type, source, author, publication date, media, attributes, and collection items.
Always retain the canonical URL from the response. AI-generated summaries should remain traceable to the public source, especially when the underlying page can change.
Boundaries
Public data only
- The server is stateless and does not keep request history.
- It cannot access private pages or bypass access controls.
- The MCP tools use the same public limits and account credits as the API.
extractor.sh does not bypass CAPTCHAs, login walls, paywalls, access controls, or regional restrictions. Review the source’s terms and applicable law before collecting or reusing data.