wigolo examples

July 17, 2026 · View on GitHub

Runnable, self-contained examples for every way agents and scripts talk to wigolo — the local-first web intelligence server. Each directory has its own README with real captured output; terminal-flavored ones ship a rendered demo.gif plus the demo.tape that produced it.

Works the same for coding agents (Claude Code, Cursor, Zed, ...) and self-hosted automation (n8n-style platforms, cron jobs, plain shell).

ExampleWhat it shows
quickstart-claude-codenpx wigolo init --agents=claude-code — one command wires Claude Code, with a per-component setup report
one-shot-clievery tool as a terminal command: search, fetch, research, and the --json | jq contract
shell-ndjson-pipelinepipe commands into wigolo shell --json — one JSON doc per line, jq filters, exit codes you can gate on
rest-curlwigolo serve + curl: /health, /v1/tools, POST /v1/search, /openapi.json, bearer-token mode
sdk-typescript-researchwigolo-sdk embedded local mode — spawn-or-reuse a daemon and run research in ~20 lines
sdk-python-agentthe wigolo PyPI client — local_client() + the autonomous agent tool with a JSON Schema
vercel-ai-sdk-toolswigolo as Vercel AI SDK tools — register webSearch / webFetch / research on any model
n8n-remote-mcpconfig reference: point self-hosted n8n (or any remote MCP/REST client) at wigolo serve
watch-changelog-webhookchange-watch jobs on a changelog, on-demand checks, diff reports, webhook delivery
plugin-search-engineminimal searchEngine plugin — extend wigolo's engine set with your own source

Previews

Set up Claude Code in one commandquickstart-claude-code:

wigolo init wiring Claude Code

The one-shot CLIone-shot-cli:

search, fetch, and json piping from the terminal

NDJSON scriptingshell-ndjson-pipeline:

piping commands through wigolo shell --json

Conventions

  • Scripts default to npx wigolo; set WIGOLO="wigolo" (or any entry point) to override.
  • Examples that talk to a daemon use loopback ports in the 3477+ range and clean up after themselves.
  • Everything runs keyless. Optional cloud or local language-model keys add answer synthesis and more schema-faithful extraction, and every example works without them.