Workflow · Level 3 — Full agentic

It watches. You get the text.

By Raj Patel

Level 3 · It comes to you

This is where the agent stops waiting for instructions and starts coming to you. Set a tripwire — a fare under $250, a price drop, a deadline approaching — and your Muse checks on a schedule and pings you only when it fires. You set it once; it watches for months. Alerts only, never purchases.

Connect these first

Connect the stores and travel sites you want watched, plus email or calendar for the alerts to reach you. Watches are read-only by design.

01

Set the tripwire

Tell Muse exactly what ‘tell me’ means: the item or flight, the price threshold, and how often to check — daily for fares, weekly for gadgets. A precise tripwire means alerts you act on; a vague one means noise you learn to ignore.

Expedia

Paste into Muse

Watch round-trip flights to [Chicago] for [Dec 18–22, 2 adults]. Text me the moment one drops under $[250]. Check 3 times a day. Do not book anything — alert only, with the price, airline, and a link.

02

Watch the price

The espresso machine, the laptop, the sold-out item — Muse checks the price on your schedule and texts you only when it drops below your threshold. You set it once and forget it; the alert arrives when there’s actually something to buy.

Walmart · Best Buy

Paste into Muse

Watch this [espresso machine] at [store link]. Alert me if the price drops below $[300] or if it comes back in stock. Check twice a day. Never purchase — just alert me with the link.

03

Watch the deadline

Renewals, bills, and dates that punish forgetfulness — Muse watches them and warns you before they hit, not after. You pick how far ahead: a week for the domain renewal, a day for the bill. Late fees become a choice, not a surprise.

Google Calendar

Paste into Muse

Remind me 3 days before [passport renewal / property tax due / subscription renewal on March 1]: what it is, what I need to do, and the link or phone number. Remind me again the day before if I haven't confirmed it's handled.

04

Keep watches few

The maintenance step: once a month, Muse lists your active watches and you kill the stale ones — the trip you already booked, the gadget you bought. Five live watches beat fifty forgotten ones; this step keeps the signal high.

Paste into Muse

List all my active watches: what each is watching, the threshold, and when it last checked. Flag any that haven't fired in 60 days so I can decide whether to keep them.

Where approvals sit

Watches never buy — that's the whole design. Alerts are informational; any purchase stays a separate, explicit instruction from you (see the shopping workflow pattern: build the cart, you check out).

What to verify

When an alert fires, open the link yourself before acting — prices move between the agent's check and your click. Once a month, glance at the active-watches list to confirm they're still checking (a broken login silently kills a watch).

What can go wrong

Two honest limits. First, stale data: the agent checks on a schedule, not continuously, so a flash sale can come and go between checks — and some sites (Amazon notably began blocking Muse in September 2026 — TechCrunch, Oct 2026) refuse agent traffic entirely. Second, alert fatigue: ten watches pinging daily trains you to ignore them. Keep it to the five that matter.

← Apply faster, still sound like you.All workflowsNo more 200-message threads. →

It watches. You get the text: questions

Standing watches raise one big question — will it buy the thing when the price drops? No: watches are alert-only by default. These answers cover that, how watches differ from price-tracking sites, and what happens when a site blocks the agent.

How is this different from a price-tracking website?
Those track one store well. Your agent watches across stores, flights, and deadlines in one place, in plain language, with thresholds you describe however you like — no app per store.
Will it buy it for me when the price drops?
Only if you separately and explicitly tell it to, with an exact price and your approval in the moment. Watches are alert-only by default — that's what makes them safe to leave running.
What if the site blocks the agent?
It should tell you the watch can't reach the site. Don't work around blocks by handing over extra credentials — pick a watch the agent can actually see, or check that one yourself.