Enables federated web search and page extraction through a self-hosted MCP/HTTP server that discovers, scores, and rotates across a fleet of public SearXNG instances, failing over to healthy ones and attributing each result to its source instance and engine. It also fetches public pages as Markdown, text, or HTML with optional headless-browser rendering, same-host crawling, PDF parsing, and redirect reporting.
Enables models to run web, image, news, places, maps, video, shopping, scholar, patent, review, lens, and page-extraction searches through Serper.dev, with per-call cost checks and a local spend ledger.