Wigolo 🔌 MCP server Open source

What it does

Wigolo is a local-first "web intelligence" MCP server that gives an AI coding agent one surface for everything web-related — search, fetch, crawl, extract, cache, find-similar, research and autonomous multi-step "gather" loops — instead of the agent shelling out to ad hoc scripts or one-off web-search tools.

Tools it gives your agent

Per its README: web search across 18 search-engine adapters, page fetch and structured extraction, whole-site crawling, a persistent local cache, similarity search over previously fetched pages, a multi-step "research" tool, and an autonomous "gather" mode. Results include verbatim excerpts pinned to byte-offset locations in the source page, aimed at reducing hallucinated citations.

Works with

Runs as an MCP server (JSON-RPC over stdio) alongside coding agents including Claude Code, Cursor, Codex, Gemini CLI, OpenCode, VS Code, Windsurf, Zed and Antigravity, and can be embedded via SDK into agent frameworks such as LangChain, CrewAI, LlamaIndex and the Vercel AI SDK.

How to install or connect

npx wigolo init
npx wigolo init --agents=claude-code,cursor

Requires Node ≥20 and about 1.5 GB of disk space; runs on macOS, Linux and Windows. All data is kept locally under ~/.wigolo/.

Maintenance and safety

AGPL-3.0-only licensed, with roughly 5.3k stars and over 2,000 commits, currently labeled a "public beta." Default operation needs no external API keys and performs searches/fetches at no per-query API cost; optional LLM-powered synthesis (used by the research/gather tools) requires an explicit opt-in and your own key for a provider like Gemini or OpenAI. The AGPL license is worth weighing for closed-source embedding.

Who should use it

Teams that want their coding agent to search and read the live web (docs, issues, changelogs) without paying per-query API fees or juggling several single-purpose search tools, and who are comfortable with the AGPL license and beta-stage maturity.