SourceWeave Web Search

Three tools for web research that run entirely on your machine. Search with SearXNG, extract clean HTML via Crawl4AI, and cache results in Redis or Valkey. Call search_web to get page_id handles, then read_pages to pull content, or skip straight to read_urls for direct fetches. The managed runtime starts a Docker Compose stack automatically when you launch it with uvx, discovers existing stacks if they're already running, and tears down containers when the last process exits. Set explicit endpoint variables if you're running the services yourself or pointing at hosted instances. The direct CLI works without MCP wrapping when you just need the search primitives. Also ships as a container image and a standalone OpenWebUI tool file.