npx sniffington: your brand mentions from the terminal

Most of the time you want one answer: did anyone say anything about us that I should read? Opening a browser tab for that felt like too much ceremony, so we wrote a command line.

Install it and sign in

If you have Node 18 or later, npx runs it without installing anything. Save an API key once and every command after that just works.

npx sniffington
sniffington login ss_sk_…
sniffington brief

The key stays at ~/.config/sniffington/config.json, mode 0600. It never leaves your machine except as the Authorization header to sniffington.com (or whatever address you point it at with --url).

Five commands, one flag

• brief prints open mentions, most urgent first, each with the one-line reason the AI gave. Add --project to narrow it to one brand.
• mentions is the feed. Filter with --status, --sentiment, --q, --range or --limit. Pipe it to grep, jq, whatever.
• sniff <brand> checks any brand on Reddit, Hacker News and Bluesky right now. Add --about to help with namesakes.
• whoami shows the workspace and plan this key reads.
• mcp prints how to connect an AI agent instead of using the CLI.

Every command takes --json. With it, the output is one JSON document instead of text. A cron job, a shell script, or another tool can read it without parsing.

Why not just use the MCP server?

The MCP server is for AI agents: Claude, ChatGPT, Cursor. The CLI is for people and scripts. You don't need an AI in the loop to ask "what needs me today" and read five lines of text. The sniff command does talk to the MCP server under the hood, because that's where the AI check runs, but the rest hits the REST API directly.

If you want both, that works too. The CLI and the MCP server use the same API keys. Set up pages for every client are under AI agents, and the CLI's own page is at /agents/cli.