Can Claude hand a shopping task to Muse?

An attendee described a Claude agent that prepares a daily to-do list. Could a birthday reminder pass to Muse to find and buy the gift?

The example was a party on Sunday for a five-year-old. The request needs a total budget and a delivery deadline.

What to try today

Use Claude to extract one task from the daily brief. Replace the bracketed fields in this prompt before using it.

Turn the birthday-gift item in my daily brief into a standalone shopping request for Muse.
The child is five. The party is on [date]. The gift must arrive by [date].
Interests: [interests]. Exclude: [restrictions].
Total budget, including tax and delivery: [amount and currency].
Delivery area: [city or postal code].
Ask me for any missing information. Do not invent preferences or a budget.
Include only the information needed for this gift. Do not include the rest of my daily brief.
Tell Muse to research suitable gifts and return product links, total prices and delivery estimates.
Require my explicit approval of the selected item and final total before purchase.
Do not send the request or buy anything.

Review the request, then paste it into your Muse conversation with the instructions below. This step assumes you already have Muse access.

Research the gift described below. Do not purchase anything yet.
[Paste the shopping request prepared in Claude.]
Check age suitability and seller availability. Include tax and delivery in the total.
If delivery by the deadline cannot be confirmed, say so.
Return your recommendation with the source product link and any unresolved details.
Stop for my approval before checkout. Do not substitute another item or increase the budget without asking.

Inspect the suggested item and delivery estimate. If you want to buy, use Muse’s supported shopping connection and review the checkout approval.

Enter credentials through the service’s own sign-in flow. Keep passwords and payment-card details out of the prompt.

The simpler future email route

Claude’s Gmail connector supports sending email and requests approval by default. Meta’s September 24 announcement describes a future email address for Muse.

Once that address becomes available, direct email is the first connection to test. There is no need to add an automation platform solely to send it.

1. Verify Muse’s official address and instructions in your account. Check whether it accepts tasks from the connected Gmail address.
2. Connect Gmail through Claude’s connector settings. Review the requested permissions and any workplace restrictions.
3. Ask Claude to draft the gift request using the prompt below. Approve the recipient and message before it sends.
4. Run a research-only test. Confirm Muse received the task and respected the spending boundary before relying on the connection.

Draft an email containing only the approved gift request to [verified Muse email address].
Include the total budget and delivery deadline. Ask Muse to research options and wait for my approval before purchasing.
Show me the recipient and the complete message before sending. Do not send until I approve.

Approval to send an email is separate from approval to purchase. An email receipt alone does not prove that Muse accepted or completed the task.

Scheduled Claude tasks need a separate account-level check for connector access. A connector working in an interactive chat does not establish unattended execution.

When Zapier or Make helps

For this example, start with the existing Claude daily brief. Adding Zapier or Make would add setup without resolving Muse’s unreleased email capability.

Use Zapier when an external event or independent schedule needs to trigger a simple email workflow. That is our implementation recommendation for this narrow job.

Make is reasonable if your team already uses its scenarios or needs custom routing. Neither platform has been tested here as a direct consumer-Muse connection.

A model API is also a different capability. Access to a Muse model does not establish control over a person’s existing Muse shopping agent.

What this teaches about agent handoffs

A useful handoff contains the task and enough context to complete it. It also names the action that requires approval.

The same pattern can turn a client-meeting brief into a research task, or a customer escalation into a proposed follow-up. Each connection needs its own test.

Sources and verification

This guide separates documented capabilities from proposed setup steps. Product guidance was checked on September 25, 2026.

• Meta: Connect 2026 announcements
• Claude: Google Workspace connectors
• Meta: How we built safety into Muse
• Zapier: Triggers and actions
• Make: Built-in connector on Claude

This question is shared under the Chatham House Rule. Participant identities and affiliations are omitted.