juanibiapina/web-search

Web Search

Multi-provider web search and content extraction via the websearch CLI.

Commands

search: Find web pages

websearch search "query"                       # Search (default: Tavily)
websearch search "query" -p brave              # Specific provider
websearch search "query" -n 10                 # More results (default: 5)
websearch search "query" --content             # Include page content
websearch search "query" --freshness week      # Filter: day, week, month, year
websearch search "query" --country DE          # Country-specific results
websearch search "query" -p serpapi --engine google_scholar  # SerpAPI engine

extract: Get page content as markdown

websearch extract "https://example.com/article"
websearch extract "https://docs.rust-lang.org/book/ch04-01-what-is-ownership.html"

Extraction is always local (fetches HTML, parses with Readability, converts to markdown). No API credits used.

answer: Direct answer with citations

websearch answer "What is the latest version of Node.js?"
websearch answer "latest React version" -p exa

similar: Find related pages

websearch similar "https://example.com/article"      # Exa only
websearch similar "https://example.com/article" -n 10

code: Find code examples

websearch code "React hooks state management"         # Exa only
websearch code "Express.js middleware authentication"

All commands support --json for raw JSON output.

Provider Guide

| Provider | Best for | Free tier | |---|---|---| | tavily | General AI-optimized search, answers | 1,000/month | | exa | Semantic search, similar pages, code context | 1,000/month | | websearchapi | Google-powered search, generous quota | 2,000/month | | brave | Independent index, privacy-focused | ~1,000/month | | serpapi | YouTube, Scholar, Amazon (40+ engines) | 100/month |

Defaults: search→tavily, extract→local, answer→tavily, similar→exa, code→exa

Provider characteristics

• tavily (default): Returns the richest snippets, noticeably more extracted content per result than other providers. Good mix of official docs and community sources. Best general-purpose choice.
• exa: Semantic/neural search that finds the freshest content and understands meaning beyond keywords. Great for "what's the latest on X" or exploratory queries. Can occasionally return lower-quality SEO-heavy content for common topics. Only provider with similar and code commands.
• websearchapi: Google-powered results. Reliably surfaces canonical/official sources (official docs, Stack Overflow). Most generous free tier (2,000/month), so it's a good fallback when other quotas run low.
• brave: Independent index (not Google/Bing). Also reliably surfaces official docs and vendor guides. Good alternative perspective to Google-based results.
• serpapi: Tight free tier (100/month) and sometimes returns fewer results than requested. Reserve for specialized engines: -p serpapi --engine youtube for video tutorials, --engine google_scholar for academic papers, --engine google_news for news.

Setup

Each provider needs an API key as an environment variable. Only configure the ones you use:

TAVILY_API_KEY      # https://app.tavily.com
EXA_API_KEY         # https://dashboard.exa.ai
WEBSEARCHAPI_KEY    # https://websearchapi.ai
BRAVE_API_KEY       # https://api-dashboard.search.brave.com
SERPAPI_KEY         # https://serpapi.com/manage-api-key

Tips

• Use --content to include page text in search results (avoids a separate extract call)
• Use answer for factual questions. Returns a direct answer with sources.
• Use similar to discover related pages from a known good URL
• Use code to find real working code examples instead of relying on training data
• Use extract to read long documentation pages (free, no API credits)
• Use -p serpapi --engine youtube to search YouTube, or --engine google_scholar for papers
• When a provider's quota runs low, switch with -p