Repolode
Submit
KnockOutEZ

KnockOutEZ/wigolo

Local-first web search, fetch, crawl and research tools for AI agents over MCP, with no API keys and no per-query cost.

Open-source alternative to:FirecrawlExaTavily

View on GitHubWebsite
Stars
5,493
Forks
449
Open issues
78
Last commit
21 hours ago

What is it?

wigolo gives an AI agent one set of web tools: multi-engine search, page fetch, multi-page crawl, structured extraction, a local cache, find-similar, cited research, and an autonomous gather loop. It runs as an MCP server next to coding agents such as Claude Code, Cursor, and VS Code, and can also be served as a REST endpoint or embedded through TypeScript and Python SDKs.

The core tools need no API keys, and cached pages and models are stored under ~/.wigolo/. Fetching escalates from plain HTTP to a headless browser when a page needs JavaScript or hits a challenge, and failed engines or blocked pages are reported in the output rather than hidden. Synthesized answers for research and agent tasks are optional and require an LLM provider, which can be a local Ollama model.

The project installs with npx wigolo init, requires Node 20 or later and about 1.5 GB of disk, and ships LangChain, CrewAI, LlamaIndex and Vercel AI SDK integrations.

Key features

  • 01Multi-engine web search with rank fusion and an explainable per-result score.
  • 02Page fetching that escalates to a headless browser for SPAs and anti-bot challenges.
  • 03Multi-page crawling via BFS, DFS, sitemap, or map-only modes with robots.txt respect.
  • 04Structured extraction of tables, metadata, JSON-LD, and custom JSON Schema.
  • 05Local cache with keyword and semantic search, plus change detection via diff and watch.
  • 06Works over MCP, a REST API, the CLI, and TypeScript and Python SDKs.

Use cases

  • Giving a coding agent live web search and page reading without paying per query.
  • Running self-hosted agents that need a web tool reachable over REST.
  • Building RAG pipelines in LangChain or LlamaIndex that load fresh web pages.
  • Monitoring a page for changes and sending updates to a webhook.

Quick start

  1. 1

    Set up the local engine

    npx wigolo init
  2. 2

    Check the setup is healthy

    npx wigolo doctor

Things to know

  • The license is listed as NOASSERTION, and the README badge says AGPL-3.0, so check LICENSING.md before commercial use.
  • Requires Node 20 or later and about 1.5 GB of free disk space for the browser engine and models.
  • Research, agent and answer synthesis need a free LLM key for a finished cited answer; otherwise they return a raw brief.
  • The project is in public beta, so expect rough edges.

Topics

#ai-agent#mcp#web-search#web-scraping#web-crawler#local-first#privacy#rag#agent#ai

FAQ

What is wigolo?
Local-first web search, fetch, crawl and research tools for AI agents over MCP, with no API keys and no per-query cost. wigolo gives an AI agent one set of web tools: multi-engine search, page fetch, multi-page crawl, structured extraction, a local cache, find-similar, cited research, and an autonomous gather loop. It runs as an MCP server next to coding agents such as Claude Code, Cursor, and VS Code, and can also be served as a REST endpoint or embedded through TypeScript and Python SDKs.
Is wigolo free to use? What license does it have?
wigolo is open source under the Other license. Check the license terms before using it in commercial products.
Is wigolo still maintained?
Actively maintained. The last commit to wigolo was on Oct 20, 2026.
What can I use wigolo for?
Giving a coding agent live web search and page reading without paying per query. Running self-hosted agents that need a web tool reachable over REST. Building RAG pipelines in LangChain or LlamaIndex that load fresh web pages. Monitoring a page for changes and sending updates to a webhook.
How do I install wigolo?
Set up the local engine: npx wigolo init Check the setup is healthy: npx wigolo doctor
Where can I get wigolo?
The source code and installation instructions are on GitHub: https://github.com/KnockOutEZ/wigolo

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects