Wigolo MCP Server

Web research in wigolo connects multi-engine search results with page retrieval, documentation crawls, and structured extraction. Search returns scored excerpts and citations; fetch can handle JavaScript-rendered pages and reuse stored login sessions. The extraction tool reads tables, metadata, brand details, or fields defined by a JSON Schema from pages or raw HTML. Fetched and crawled content remains in a persistent local cache for later searches without network access. Related-page discovery combines keyword matches, on-device embeddings, and optional live web results, with cold-start warnings when local evidence is sparse.

For broader questions, research breaks the task into subqueries and gathers sources, returning a cited report when synthesis is configured or a structured brief otherwise. The agent tool accepts a natural-language gathering request, can extract schema-defined fields, and records its steps while working within time and page budgets. Core retrieval tools need no API keys; optional synthesis can use a configured provider or a local model. For follow-up checks, diff compares Markdown or cached pages, while watch retains monitoring jobs across sessions. Watch checks require an explicit request or another wigolo tool call after the interval has elapsed; an idle server does not poll independently.