Let your AI agent watch a shop for you

Scout watches public store collection URLs and emails you when new products appear — hourly to weekly, with a baseline first scan and per-watch pause.

You already ask AI agents to do chores. This page explains what happens when the chore is “keep an eye on this store’s new-arrivals page and tell me what shows up” — and why Scout is built so an agent can do it safely.

The agent needs credentials that last. Browser sessions expire in about an hour. An agent that has to re-login every day is an agent you will turn off. Scout issues agent keys — long-lived sk_scout_… Bearer credentials created in the web app, shown once, stored hashed, revocable any time. Hand one to your agent and it stays signed in for as long as the key lives.

The agent gets the same surface as you. Scout’s hosted MCP server at https://scout.clarved.com/mcp exposes 7 tools — list, add, update, delete, inspect, scan now, and account status. Under the hood they are the same functions the web dashboard calls, so the same 10-watch limit, the same cadence menu (hourly to weekly), and the same honest errors apply. An agent cannot exceed the plan or reach a privileged path.

Monitoring stays polite. Whether a human clicks “Check now” or an agent calls trigger_scan, Scout reads only the one public collection URL on the watch, on the cadence you picked, and never logs in or bypasses a paywall. Agents make this easier to keep polite, not harder: the cadence is explicit, and pausing is one tool call away.

Emails still come to you. The digest of new products is emailed to the account owner, not to the agent. The agent can read recent results through Scout’s tools, but the notification channel stays yours. Manual agent scans don’t email at all; scheduled scans do.

Ready? Connect Scout to your AI agent has the five-minute setup with copy-paste blocks for Claude Code, Cursor, Codex, VS Code, and Gemini CLI.