Replyz / Guides
Guides: email for AI agents, step by step.
Short how-tos for people who run personal agents. Each one shows the real commands and says what Replyz can and cannot do. Replyz has no cold sending: an agent can reply only to people who emailed it.
claude mcp add
Add a remote MCP server to Claude Code, send an API key as a header and check the connection.
Get a verification code from email
Wait for the email and read the code or link in one call, from the CLI, MCP or REST.
Create an email address for an AI agent
How an agent signs itself up and how a human claims the account with one email.
Wait for an email in code
Polling, webhooks or a long poll, and how the wait call and the event feed work.
Prompt injection in email
What a mail service can reduce and what the agent itself must do.
Email MCP servers compared
Four servers from their own documentation: Gmail, Resend, AgentMail and Replyz.
Where your agent's email is stored
What is stored in the EU, what is processed globally, and how long mail is kept.
Keep an API key out of the agent's context
Let a program read the key, not the model, and know what is still exposed.
Why the email never arrived
A checklist for when the verification email does not reach the agent's inbox.
Claim your agent's account
How you take ownership of the account your agent created, with one email.
Is receiving enough?
Which agent jobs only need to receive, and what Replyz does today.
Looking for a specific agent?
See Replyz for AI agents for Claude Code, Codex, OpenClaw, Hermes, Cursor and others, and the Replyz MCP server and the CLI for every tool and command.