AI drafts instant replies to common inquiries. Your team reviews and approves before delivery — maintaining quality at 10x the speed.
Support teams spend over 60% of their time answering repetitive inquiries — password resets, billing questions, shipping status updates. This creates long response times that frustrate customers and burn out agents. Meanwhile, the nuanced tickets that actually need human expertise get buried in the queue.
MultiMail's AI agent reads incoming support emails, matches them against your knowledge base, and drafts accurate replies instantly. With gated_send oversight, every response is queued for human approval before delivery — so your team gets speed without sacrificing accuracy or brand voice.
Create a MultiMail mailbox for your support address. Incoming emails are processed in real time and made available to your AI agent via the API or MCP tools.
Your agent uses check_inbox and read_email to pull new messages, then classifies each inquiry by topic (billing, shipping, account, technical) and matches it to your response templates or knowledge base.
The agent composes a personalized reply using context from the customer's email, account history, and your approved response patterns. Drafts are submitted via the send_email or reply_email tool.
With gated_send oversight, drafted replies land in an approval queue. Your support team reviews each response, edits if needed, and approves with one click. Approved emails are sent immediately.
Track approval rates and edit frequency. As your team builds confidence in the AI's drafts, you can adjust oversight modes per category — moving low-risk replies to monitored while keeping sensitive topics gated.
import requests
API = "https://api.multimail.dev/v1"
HEADERS = {"Authorization": "Bearer mm_live_xxx"}
"cm"># Check for new emails in the support mailbox
response = requests.get(
f"{API}/mailboxes/support-mailbox-id/inbox",
headers=HEADERS,
params={"status": "unread", "limit": 20}
)
inbox = response.json()
for message in inbox["emails"]:
"cm"># Read full email content
email = requests.get(
f"{API}/emails/{message[&"cm">#039;id']}",
headers=HEADERS
).json()
print(f"From: {email[&"cm">#039;from']} Subject: {email['subject']}")Check your support mailbox for unread messages and retrieve full email content for processing.
import requests
API = "https://api.multimail.dev/v1"
HEADERS = {"Authorization": "Bearer mm_live_xxx"}
def draft_support_reply(email: dict) -> str:
"""Your AI classification + drafting logic."""
"cm"># Classify: billing, shipping, account, technical
category = classify_inquiry(email["text_body"])
"cm"># Generate reply from knowledge base
return generate_reply(category, email)
"cm"># Draft and send reply with gated oversight
draft = draft_support_reply(email)
response = requests.post(
f"{API}/reply",
headers=HEADERS,
json={
"in_reply_to": email["id"],
"text_body": draft,
"html_body": f"<p>{draft}</p>"
}
)
print(f"Reply queued for approval: {response.json()[&"cm">#039;id']}")Use your AI model to classify the inquiry and generate a draft reply, then send it through MultiMail with gated oversight.
"cm">// Using MultiMail MCP tools in an AI agent
"cm">// 1. Check for new support emails
const inbox = await mcp.check_inbox({
mailbox_id: "support-mailbox-id",
status: "unread",
limit: 10
});
"cm">// 2. Read the full email
const email = await mcp.read_email({
email_id: inbox.emails[0].id
});
"cm">// 3. Tag it for tracking
await mcp.tag_email({
email_id: email.id,
tags: ["auto-reply", "billing"]
});
"cm">// 4. Send reply (goes to approval queue under gated_send)
await mcp.reply_email({
email_id: email.id,
text_body: "Hi! Your billing cycle resets on the 1st..."
});If your agent connects via MCP, use the built-in tools to check inbox and reply directly.
AI drafts replies in seconds. Your team only needs to review and approve, cutting average response time from hours to minutes.
Every reply follows your approved templates and tone guidelines. No more inconsistency between agents or shifts.
Gated send oversight ensures no email reaches a customer without human approval. Build trust gradually and adjust oversight as confidence grows.
When AI handles the repetitive 60%, your support agents can focus on the nuanced tickets that actually need human judgment and empathy.
Every draft, approval, and sent email is logged. Track which responses were edited, measure AI accuracy, and identify areas for improvement.
Email infrastructure built for AI agents. Verifiable identity, graduated oversight, and a 38-tool MCP server. Formally verified in Lean 4.