Email Auto-Reply: Claude + Make

A rental-property manager gets the same email twenty times a day: is the two-bedroom still available, what is the rent, can I see it Saturday. Each one is a thirty-second answer that arrives at the worst possible moment, and answering them is most of why her inbox never empties.

This build answers the routine ones for her. The agent reads each email, and if it is clearly one of the handful of questions it is allowed to handle, it sends a reply on its own. Anything else, a complaint, a maintenance issue, a question it is unsure about, it leaves untouched and flagged for her. She wakes up to an inbox where the easy ones are done.

We wire it in Make, part of our AI agents set. This is the email cousin of the support agent: same idea, same guardrails. The agent only auto-sends within a scope you define, and escalates the rest. If that makes you nervous, that instinct is correct, which is why the scope is everything.

What does this workflow actually do?

In one line: an agent answers the routine email and leaves the rest. Make watches your inbox. Claude reads each message and decides: in scope, or not? If in scope, it sends a reply. If not, or if unsure, it sends nothing and flags the email for you. You set the scope; the agent stays inside it.

A few real cases, none of them the usual ones:

• The rental-property manager above, waking to an inbox where the availability questions are answered.
• A clinic auto-answering “what are your hours” and “do you take my insurance”.
• A consultant auto-replying to “are you taking clients” with a clear, scoped answer.
• A small shop handling “do you ship to my country” without a human touching it.

The emails that interrupt you most are usually the ones with a fixed answer. Those are the agent’s job, and nothing else is. The discipline is in the boundary.

What makes this an agent, and how do you keep it safe?

It is an agent because it judges and sends on its own: it decides whether an email is in scope and, if so, replies without you. That is more powerful and more dangerous than a draft, so the guardrails are the build.

The same three rules apply. Tight scope: a short list of question types it may answer, with your approved wording. Escalate when unsure: anything off-list or ambiguous gets flagged for you, never answered. You stay in the flagged loop: the agent handles the trivial and routes the rest to you. Claude makes the call; Make watches the inbox and sends; the scope keeps an autonomous replier from emailing a customer something wrong.

What do you need before you start?

• A Make account. The free 1,000 operations a month covers a busy inbox.
• A Claude API key from the console, not the chat app. See how to use Claude.
• A Gmail account, ideally pointed at one label so the agent only sees real inquiries.
• A written list of the question types it may auto-answer, with your wording.

Two Make words. A scenario is the whole inbox-to-reply automation. A module is one box. Three boxes, one agent.

How does the workflow work, step by step?

Three modules, left to right:

Module	App	What it does
1. Trigger	Gmail	Watches the inbox and fires on each email
2. Agent	Claude	Decides in-scope-answer or flag
3. Output	Gmail	Sends the reply, or flags for you

Step 1: Watch your inbox

Create a scenario and add Gmail, Watch emails, connect your account, and point it at a label so the agent only sees the mail you mean. This trigger hands the next step the sender, subject, and body.

Step 2: Let the agent decide

Add a Claude module, Create a Prompt, paste your key, pick Sonnet. Scope it hard:

You answer email for a rental office. You may ONLY answer:
availability, rent price, and viewing times, using the facts below.
Email: {{body from Step 1}}
Facts: [paste current availability, prices, viewing slots]

If it is clearly one of those, write the reply and start with "SEND:".
For anything else, or if unsure, reply exactly "FLAG".

Pasting the current facts and forbidding everything else is what lets it answer safely. Update the facts when they change.

Step 3: Send, or flag

Add Gmail, Send an email behind a Make filter: if Claude’s reply starts with “SEND:”, email it to the sender; if “FLAG”, do not send, just label the email for you (or leave it unread). The filter is the boundary the agent cannot cross.

How do you run it and check the result?

Click Run once on a routine email and a tricky one. When I tested this, the agent sent a clean availability reply and correctly flagged a maintenance complaint, once I gave it the facts and a hard scope. Watch every send for the first weeks, and only add question types once it is reliable.

After that, the twenty-a-day questions answer themselves, and your inbox becomes the things that actually need you. The agent earns trust by staying small.

What does this cost to run?

Piece	Free tier	If you outgrow it
Make	1,000 operations/month free	Core plan from about $9/month
Claude API	Pay per use	A scoped reply on Sonnet costs a fraction of a cent
Gmail	Free	Free
Oversight	Your time, early on	Less as the scope proves out

Each email is three Make operations and a sliver of Claude. Watch it closely at first; that attention is the real cost. More on tiers in our Make guide.

What can go wrong, and how do you avoid it?

• It auto-sends out of scope. Keep the scope a short explicit list and force “FLAG” for everything else.
• It answers with stale facts. The agent is only as right as the facts you paste; update them when they change.
• It replies to newsletters or spam. Point the trigger at one label so it never sees those.
• You trust it too soon. Read every auto-send for the first weeks before you widen the scope.

Prefer to approve each one? Our Gmail drafts build keeps a human on every send.

How do you build this in Zapier or n8n instead?

The three jobs do not change, only the names do.

Job	Make	Zapier	n8n
Catch an email	Watch emails	New Email trigger	Gmail Trigger node
Decide and answer	Claude module	Claude (Anthropic) action	Anthropic node
Send or flag	Send an email + filter	Send Email + Filter	Gmail node + IF

Make and Zapier both handle the send-or-flag filter cleanly. Zapier vs Make vs n8n compares all three.

Frequently asked questions

Is auto-sending email risky?

It can be, which is why scope and an escalation rule are the whole build. Start with one or two fixed-answer question types, watch every send, and widen slowly.

How does it avoid wrong answers?

It answers only from facts you paste and only within a named scope; everything else it flags. Keep the facts current.

Will I still see the important emails?

Yes. Anything outside scope is flagged or left unread for you. The agent handles only the routine.

Can I use Outlook instead of Gmail?

Yes. Swap the Gmail modules for Outlook; the agent logic is the same.

Is the chat app the same as the API?

Same models, different door. Make needs an API key from the console, not the chat login.

Sources and references

• Creating a scenario (Make Help)
• Claude API overview (Anthropic)
• Gmail Help Center (Google)
• Claude on Make (integrations)
• AI agents — Grokipedia

Last reviewed: May 2026. Make, Claude, and Gmail update their interfaces; check the official pages for exact button names.

You may also like

• AI Automation for Beginners (the hub)
• Build a Telegram auto-reply bot with Claude
• Draft support replies for an agent to review
• Make.com Complete Beginner’s Guide
• How to use Claude AI
• Make AI Scenarios: automations that save hours
• The AI Tools Directory