What are AI agents?

Heron can give an AI assistant like Claude access to your email. You ask for what you need in plain language, and it uses Heron to read, search, organize or draft for you. You decide, mailbox by mailbox, what it’s allowed to do.

A few examples

• “Give me my morning email brief.”
• “Who hasn’t replied to me this week?”
• “Draft a reply to Camille confirming Thursday.”
• “Archive the newsletters older than a week.”
• “Summarize this thread before the meeting.”

Which tools?

Everything goes through the hosted Heron server: the agent doesn’t need any local app.

• claude.ai, Claude Desktop and the Claude mobile app, through a custom connector: see Connect Claude.
• Claude Code, with one command: see Claude Code.
• Any agent that can run a command, with the heron CLI: see The heron CLI.

Server address (MCP protocol):

https://api.heronmailapp.com/mcp

You stay in control

• By default, AI agent access is off and every mailbox is set to No access.
• You pick a level per mailbox: read, drafts and organizing, or sending. See Enable access and choose levels.
• With Confirm before sending (on by default), nothing is sent without your approval: see Approve an agent’s sends.
• Every action is recorded in the activity log, and you can revoke access at any time.

Good to know: what an agent reads in your email is sent to its AI provider (Anthropic for Claude). Only give access to the mailboxes it needs. Also read Security best practices.

For a fuller overview, see the Heron + Claude page.