sarvcrawl

Description

SarvCrawl turns websites and documents into a knowledge base your AI client can query. The MCP server exposes 21 tools: scrape a single page, crawl a whole site, map a site's URLs, run a web search, then search everything you ingested with BM25 keyword search or hybrid semantic search. Crawls run as background jobs. The tool descriptions tell the model to start the job, poll its status and then search the results, so an agent can do the whole crawl-then-answer loop from one prompt without glue code. Pages are converted to clean markdown. PDFs and Office files are parsed, with OCR for scanned pages. Content is chunked and embedded for you, so you don't need your own vector database. Use it over stdio with npx -y sarvcrawl-mcp, or connect to the hosted endpoint at https://crawl.sarv.com/mcp from ChatGPT and other remote MCP clients. Sign-up includes 1 billion free tokens and needs no card.

Community Metrics

Categories

• mcp

Tags

• mcp
• smithery

Capabilities

• Remote MCP server

Information