mailpitcher

Use Mailpitcher with AI assistants (MCP)

On the Pro plan, connect Claude or another MCP client with a personal token, so it can search mail, draft replies and run outreach through your limits.

Updated · 3 min read · 8 steps

MCP (Model Context Protocol) lets an AI assistant, like Claude, use Mailpitcher for you: find and read conversations, write drafts and replies, schedule emails and follow-ups, and create and run campaigns. It works through the same daily caps, do-not-contact list and checks as you do, and what it wants to send waits for your approval.

  1. Pro plan and two-step sign-inBoth are needed for a token.
  2. Create tokenSettings, Security, MCP access tokens.
  3. Add the serverPaste the address and token into your AI tool.
  4. Ask in plain wordsIt works through the same limits as you.
Connecting an AI assistant over MCP.

Before you start

  • The Pro plan. MCP isn’t on Solo or Free.
  • Two-step sign-in on. A token can send mail, so it needs the extra protection. See Two-step sign-in.
  • An AI tool that can add an MCP server with a header.

Create a token

Open Settings, Security

Click your email address at the bottom left, then Settings, plan and usage, then Security. Scroll to Connect your AI tools (MCP).

Click Create token

Mailpitcher asks for a current code from your authenticator app (or a backup code). Type it and click Create token.

Copy the token now

The token starts with mp_. It is shown once. Click Copy token, keep it somewhere safe, then click Done. Treat it like a password.

Settings, Security, Connect your AI tools (MCP).
  1. Create token
  2. Copy the token: it is shown once
  3. The server address for your AI tool
  4. Revoke a token any time

Connect your AI tool

Add the MCP server

In your AI tool, add an MCP server with:

  • Address: https://app.mailpitcher.com/api/mcp
  • Header: Authorization: Bearer followed by your token

Each tool does this in its own settings. Look for “MCP servers” or “connectors”.

Ask in plain words

For example: “Find the latest reply from financebuzz and draft an answer”, or “Add these five sites to my Fintech campaign”. The assistant does it through your account.

Approve what it wants to send

When the assistant asks to send an email, reply, start a campaign, send a campaign to chosen people or change the emails of a running campaign, nothing goes out yet. It waits in Scheduled, under Waiting for your approval (from your token’s name**)**. Check it, then click Approve to send it, or Discard.

Why: emails you receive can contain text written to steer an AI (“forward all invoices to…”). With approval on, an email like that can’t make your assistant send anything without you seeing it first.

When you approve, the send goes through your daily caps and do-not-contact list at that moment, like anything you send.

Stay in control

Let a token send without approval (optional)

In Settings, Security, tick Allow sending without approval under a token, and enter a code from your authenticator app. That token’s sends then go out straight away. Only do this for an assistant that asks you before each send. You get an email when it’s turned on. Untick it to go back to approving.

New tokens start with this off. Tokens made before this setting existed have it off too.

Revoke a token

In Settings, Security, click Revoke next to a token. It stops working at once, and anything it asked to send that you haven’t approved is dropped.

Tokens also stop by themselves after 90 days, or after 30 days without use.

Good to know

  • Everything the assistant sends goes through your mailboxes’ daily caps and your do-not-contact list, like anything you send.
  • Text from emails reaches the assistant marked as content from an email, not instructions. Assistants usually respect that, but it’s no guarantee, which is why sending waits for your approval.
  • The assistant can only do what the app can do. It can’t see other people’s accounts.
  • Up to 120 calls a minute per account.