Scout CLI

Run Scout from the terminal, and wire the Scout MCP server into your coding agent with one command.

npm install -g @scout-ai/cli
# or run without installing: npx @scout-ai/cli search "..."

Set your key once (get one at platform.usescout.sh/settings):

export SCOUT_API_KEY=sk_your_key

Commands

scout search "best climate tech startups 2026" --limit 5 --country us
scout scrape https://example.com
scout extract https://example.com/pricing --objective "plan names and prices"
scout crawl https://example.com --max-pages 50
scout screenshot https://example.com --full-page
scout company stripe.com
scout answer https://example.com "what does this company sell?"
scout find-all "Series A fintech companies in Europe" --limit 25

Output is JSON on stdout, so it pipes into jq and friends:

scout search "serp api" --limit 3 | jq '.results[].url'

Wire it into your coding agent

One command adds the Scout MCP server (web search, scraping, extraction) to your agent's config:

scout mcp install --client claude-code
scout mcp install --client codex
scout mcp install --client gemini-cli
scout mcp install --client antigravity
scout mcp install --client cursor
scout mcp install --client windsurf

It writes the config file each client expects (e.g. .mcp.json for Claude Code, ~/.codex/config.toml for Codex, ~/.gemini/settings.json for Gemini CLI). Pass --api-key to bake the key in, or set SCOUT_API_KEY first.

To see the raw config block without writing anything:

scout mcp print

Flags

Flag	Use
--api-key <key>	Override SCOUT_API_KEY
--limit <n>	Max results (search, find-all)
--country <cc>	Two-letter country code (search)
--depth <d>	shallow | standard | deep (search)
--version, --help	Version / usage

License

MIT