OpenClaw Guide

Giving your agent an inbox enables it to:

• Sign up for its own accounts to GitHub, hosting platforms, and more, so you don't need to share your own credentials.
• Process attachments (such as receipts and invoices) and act accordingly
• Receive newsletters, parse them, and send the most important information to you
• Send you daily reports and digests
• Send and receive emails

Add the remote MCP server, then sign in. This is the same setup as the OpenClaw page.

openclaw mcp add resend --url https://mcp.resend.com/mcp --transport streamable-http
openclaw mcp login resend

Ask your agent to install the skills:

Let's get you set up with an email inbox! Install the resend-skills from https://github.com/resend/resend-skills and review them before continuing.

Your agent may also benefit from installing the Resend CLI.

Your agent will ask you to create an API key.

We recommend not pasting API keys into the chat. Instead, use one of these two methods:

1. SSH into your agent's machine and store the API key in an .env file.
2. Store the API key in a password manager like 1Password, and give your agent access to its own vault. This can be done using a 1Password Service Account on team plans.

Next, your agent needs to know the email address it will use to send and receive emails. Strongly consider using a subdomain (agent.example.com) instead of the root domain (example.com), especially if you want to receive emails.

You can also use the Resend CLI to verify a domain, although you will need to manually add the DNS records to your DNS provider.

Once your domain is verified in Resend, you can use it to send (and receive) emails with your agent. Add the email address to your agent's memory and tell it to send you a test email to your own email address.

At this point, your agent can send emails, but it can't receive emails yet.

To receive, you need to set up a webhook.

Your webhook server might look something like this:

import { Resend } from 'resend';

const resend = new Resend(process.env.RESEND_API_KEY);

// Security: strict allowlist
const ALLOWED_SENDERS = ['your@email.com'];

async function handler(req) {
  const payload = await req.text();

  const id = req.headers.get('svix-id');
  const timestamp = req.headers.get('svix-timestamp');
  const signature = req.headers.get('svix-signature');

  if (!id || !timestamp || !signature) {
    return new Response('Missing headers', { status: 400 });
  }

  const event = resend.webhooks.verify({
    payload,
    headers: { id, timestamp, signature },
    webhookSecret: process.env.RESEND_WEBHOOK_SECRET,
  });

  if (event.type === 'email.received') {
    // Security check
    if (!ALLOWED_SENDERS.includes(event.data.from.toLowerCase())) {
      return new Response('OK', { status: 200 });
    }

    // Get full email
    const { data: email } = await resend.emails.receiving.get(
      event.data.email_id,
    );

    // Notify user instantly
    await notifyUser(email);
  }

  return new Response('OK', { status: 200 });
}

One of the benefits of using Resend over other tools is that you don't need to constantly ask your agent to check its inbox, or rely on cron jobs to check every so often. With Resend, you can use OpenClaw's Gateway API to be notified instantly when your agent receives an email.

Giving your agent an inbox is incredibly powerful, but it also comes with some security concerns. The risk of prompt injection via email is a real concern.

The Resend Skill includes security guidelines. Use the leveled security approach below to decide which security level is right for your use case.

1. Strict Allowlist: Only allow emails from specific senders. Recommended for most use cases.
2. Domain Allowlist: Allow emails from any sender from a given domain (e.g. anyone from example.com).
3. Content Filtering with Sanitization: Accept emails from anyone, but sanitize content to remove potential injection attempts.
4. Sandboxed Processing: Process all emails but in a restricted context where the agent has limited capabilities.
5. Human-in-the-Loop: Process all emails but require human approval for each email.

In general, we recommend starting with the Strict Allowlist and gradually decreasing security if needed. Review additional security best practices in the Resend Skill.

If you have any questions, please contact support.