sx

Multi-engine web search from the command line

sx is a CLI tool for searching the web from your terminal. It supports multiple search backends -- SearXNG (self-hosted), Exa, Jina, Brave Search and Tavily -- with automatic fallback when the primary engine is unreachable.

This is a Go port of the original Python searxngr project, extended with multi-engine support.

Key Features

• Multiple search backends - SearXNG, Exa, Jina, Brave Search, Tavily with automatic fallback
• Multi-instance SearXNG failover - ordered or parallel-fastest strategy
• Terminal-based interface with colorized output
• Non-interactive by default for scripting; -i for interactive mode
• Search engine selection (bing, duckduckgo, google, etc. via SearXNG)
• Support for search categories (general, news, images, videos, science, etc.)
• Safe search filtering (none, moderate, strict)
• Time-range filtering (day, week, month, year)
• JSON output for scripting
• Built-in content extraction - fetch and convert results to clean markdown
• Anti-bot detection - rotating user agents, realistic headers, random delays
• Query history - searchable history with sx history
• Shell completions - bash, zsh, fish, powershell
• Cross-platform (macOS, Linux, Windows)

Installation

go install github.com/your-repo/sx@latest

Or build from source:

git clone https://github.com/your-repo/sx.git
cd sx
go build -o sx .

Configuration

Config is stored at $XDG_CONFIG_HOME/sx/config.toml (typically ~/.config/sx/config.toml). Created automatically on first run.

Example config.toml

# sx configuration file

# Primary search engine (searxng, brave, tavily, exa, jina)
engine = "searxng"

# Fallback engines tried in order if primary fails
fallback_engines = ["exa", "jina", "brave", "tavily"]

# SearXNG instance settings
searxng_url = "https://searxng.example.com"
searxng_urls = ["https://searxng-backup-1.example.com", "https://searxng-backup-2.example.com"]
searxng_strategy = "ordered" # ordered or parallel-fastest
# searxng_username = ""
# searxng_password = ""

# General settings
result_count = 10
safe_search = "strict"
http_method = "GET"
timeout = 30.0
expand = false
no_verify_ssl = false
no_user_agent = false
no_color = false
debug = false

# Output defaults
# default_output = ""       # "interactive" to default to interactive mode
history_enabled = true
max_history = 100

# Brave Search API (https://api.search.brave.com/)
# Free tier: 2,000 requests/month
[engines_brave]
api_key = ""  # or set BRAVE_API_KEY env var

# Tavily Search API (https://tavily.com/)
# Free tier: 1,000 credits/month
[engines_tavily]
api_key = ""                  # or set TAVILY_API_KEY env var
search_depth = "basic"        # basic (1 credit) or advanced (2 credits)
include_raw_content = false   # return full page content with results
include_answer = false        # return a direct answer

# Exa Search (API + MCP)
[engines_exa]
mode = "auto"                # auto, api, mcp
api_key = ""                 # or set EXA_API_KEY env var
mcp_url = ""                 # optional MCP HTTP endpoint
mcp_tool = "exa-web-search"  # MCP tool name
num_results = 10

# Jina Search
[engines_jina]
api_key = ""                 # or set JINA_API_KEY env var
allow_keyless = true
base_url = "https://s.jina.ai"

API Keys via Environment Variables

export BRAVE_API_KEY="your-brave-key"
export TAVILY_API_KEY="tvly-your-tavily-key"
export EXA_API_KEY="your-exa-key"
export JINA_API_KEY="your-jina-key"

Usage

Basic Search

sx "why is the sky blue"
sx "golang tutorials" -n 5

Select Search Engine

# Use a specific backend
sx "query" --engine exa
sx "query" --engine jina
sx "query" --engine brave
sx "query" --engine tavily

# Default: uses primary engine with automatic fallback
sx "query"

Output Links for Piping

# Get URLs only (one per line)
sx "golang testing" -L -n 5

# Pipe to other tools
sx "rust tutorials" -L -n 3 | xargs open

Fetch and Convert Pages to Markdown

# Top result as markdown
sx "golang channels tutorial" --text --top

# Multiple results saved to file
sx "rust ownership" --text -n 3 -o results.md

Pipelines with scrpr

sx pairs with scrpr for content extraction:

# Search + extract content
sx "query" -L -n 5 | scrpr --format markdown

# Save to directory
sx "query" -L -n 5 | scrpr --format markdown -o articles/

# Use Jina Reader for JS-heavy sites
sx "query" -L -n 5 | scrpr -B jina --format markdown

# With rate limiting
sx "query" -L -n 10 | scrpr --delay 0.5 --continue-on-error

Other Options

# Categories
sx "query" -N              # news
sx "query" -V              # videos
sx "query" -S              # social media
sx "query" -F              # files

# Filtering
sx "query" -r week         # time range: day, week, month, year
sx "query" -w example.com  # site-specific search
sx "query" --safe-search none

# Output formats
sx "query" --json          # JSON output
sx "query" --json -c       # Clean JSON (no null fields)
sx "query" -H              # Raw HTML with anti-bot headers

# Interactive mode
sx "query" -i

# History
sx history
sx history clear
sx history -n 50

# Shell completions
sx completion bash
sx completion zsh

All Flags

Flags:
      --categories strings   search categories (general, news, videos, images, music, etc.)
      --clean                omit empty/null values in JSON output
      --debug                show debug output
  -e, --engines strings      SearXNG engines to use
      --engine string        search backend (searxng, brave, tavily, exa, jina)
  -x, --expand               show full URLs in results
  -F, --files                files category shortcut
  -j, --first                open first result in browser
  -h, --help                 help for sx
  -H, --html                 fetch raw HTML with anti-bot headers
      --http-method string   GET or POST for SearXNG (default "GET")
  -i, --interactive          enter interactive mode after results
      --json                 JSON output
  -l, --language string      search language
  -L, --links-only           output URLs only, one per line
      --lucky                open random result in browser
  -M, --music                music category shortcut
  -N, --news                 news category shortcut
      --no-verify-ssl        skip SSL verification
      --nocolor              disable colors
      --noua                 disable user agent
  -n, --num int              results per page (default 10)
  -o, --output string        save output to file
      --safe-search string      none, moderate, strict (default "strict")
      --searxng-strategy string SearXNG instance strategy (ordered, parallel-fastest)
      --searxng-url string      Primary SearXNG instance URL
      --searxng-urls strings    Additional SearXNG instance URLs for failover
  -w, --site string             search within a specific site
  -S, --social               social media category shortcut
  -T, --text                 fetch pages and convert to markdown
  -r, --time-range string    day, week, month, year
      --timeout float        request timeout in seconds (default 30)
      --top                  show only top result
      --unsafe               disable safe search
  -v, --version              version
  -V, --videos               videos category shortcut

Search Backend Comparison

Backend	Auth	Free Tier	Best For
SearXNG	None (self-hosted)	Unlimited	Privacy, full control
Exa	API key or MCP	Varies by plan/MCP setup	Research-focused search, MCP workflows
Jina	Optional API key	Keyless best-effort	Fallback without mandatory key
Brave	API key	2,000 req/month	Fallback, quick setup
Tavily	API key	1,000 credits/month	LLM workflows, rich content

Troubleshooting

Error: all backends failed Check your primary engine URL and API keys. Use --debug for details.

Error: HTTP 429 Too Many Requests SearXNG rate limiting. Update server limiter settings or use a fallback engine.

Error: failed to parse JSON response Enable JSON format in SearXNG's settings.yml:

search:
  formats:
    - html
    - json

Dependencies

• cobra - CLI framework
• toml - Configuration
• color - Terminal colors
• go-readability - Content extraction
• html-to-markdown - HTML to Markdown