wigolo documentation

wigolo is a local-first web-intelligence server for AI agents: search, fetch, crawl, extract, research, and a persistent knowledge cache, all running on your machine with no API keys required for core work. It speaks MCP over stdio for local coding agents (Claude Code, Cursor, and friends), exposes a remote MCP endpoint and a REST API for self-hosted agents (n8n, VPS automations, your own services), ships SDKs for agent frameworks, and doubles as a plain CLI for scripts.

Everything the server returns is transparent: per-result evidence scores, per-engine telemetry, honest failure labels. Everything it stores stays in ~/.wigolo on your disk.

Pages

PageWhat's in it
Getting startedThe 5-minute path: npx wigolo init, wire an agent, run your first search.
InstallationEvery install channel — npm, Docker, MCP bundle, registries — plus the agent auto-wire matrix and uninstall.
ConfigurationResolution order, the settings TUI, and grouped env-var tables for search, fetch, models, LLM providers, cache, and serve.
ToolsThe 10 tools with parameters, response fields, and worked examples.
CLIFull command reference: management commands, one-shot tools, the interactive shell, and the --json contract.
REST APIwigolo serve, endpoints, the fail-closed auth model, resource limits, and a live curl quickstart.
SDKsTypeScript and Python clients, plus LangChain, CrewAI, LlamaIndex, and Vercel AI SDK integrations.
Self-hostingRunning wigolo where your agents run: VPS, Docker, tokens, reverse proxies, and honest notes on datacenter IPs.
SkillsAgent skill packs: the 11-pack catalog, install scopes, and the receipts model.
PluginsExtending wigolo with your own search engines and content extractors.
TroubleshootingSymptom-to-fix table, platform notes, and the FAQ.
Privacy & securityWhat lives on disk, what leaves your machine, credential handling, and responsible disclosure.

Getting help

Next: Getting started

Edit this page on GitHub