A local-first MCP and REST web intelligence layer for AI coding agents, providing keyless multi-engine search, fetching, crawling, and caching without per-query costs. It runs ranking and embeddings on-device while returning verbatim excerpts pinned to byte-exact source spans.
Project overview
Enables self-hosted agents and developers to access multi-engine web search with explainable scoring and byte-exact evidence without relying on metered APIs.
Project type
MCP · AI Agent · AI Coding · RAG
Use cases
Knowledge Q&A · Coding & Development
Deployment
Refer to project documentation
License
License pending
Best for
Developers building AI agents who need a local web intelligence tool with explainable scoring and byte-exact citations.
Users seeking a self-hosted search and fetch layer that runs without metered API keys for core operations.
Key capabilities
Executes multi-engine web search with 18 direct adapters, rank fusion, ML reranking, and an explainable per-result score.
Limitations and risks
Some challenge-protected sites score IP reputation, meaning a datacenter IP may fail where a home connection would succeed.
Without an optional LLM provider, research and agent tools return raw briefs and evidence instead of synthesized answers.
Public search engines may block or rot, though this is mitigated by engine fusion and aggregators.
Getting started
Setup is rated easy, featuring an unattended one-command init process via 'npx wigolo init', followed by 'npx wigolo doctor' to verify configuration. An optional LLM provider can be configured for synthesis features, and a GPU is optional.
Alternatives and comparisons
Provides a suite of web interaction tools for MCP clients including scraping, searching, crawling, and browser automation.
Bridges AI assistants to search capabilities via the Model Context Protocol for web search and multi-step research.