Replyz / Agents

Email for every AI agent you run.

Each agent gets an inbox of its own: it signs up for services, waits for email, reads verification codes and links, and asks you before anything is sent. Pick your agent to see how to connect it.

Agents on your machine: skill and CLI

These agents can run commands, so they use the Replyz skill and CLI. It's faster than MCP and keeps the API key in an environment variable.

OpenClaw

Add the skill, then message your agent from Telegram or WhatsApp.

Hermes Agent

hermes skills install, or MCP through config.yaml.

Claude Code

The skill in ~/.claude/skills, or claude mcp add.

Codex

The MCP server in ~/.codex/config.toml, with the key from the environment.

Agents in an app or browser: MCP

These connect to the remote MCP server at https://mcp.replyz.io/mcp with your API key as a header.

Claude

A custom connector in Claude, Claude Desktop or Cowork, with a request header.

Cursor

.cursor/mcp.json with ${env:REPLYZ_API_KEY}.

Manus

A custom MCP server in Settings → Integrations, with a Bearer token.

Gemini CLI and Antigravity

settings.json or mcp_config.json, with an Authorization header.

Any MCP client

Streamable HTTP and an Authorization header. Tool list on the MCP page.

ChatGPT

Not yet: ChatGPT connectors need OAuth, planned after launch.

Building your own agent?

Use the REST API or the JavaScript and Python SDKs. Every primitive the CLI and MCP have (create an inbox, wait, extract the code) is one API call. See the CLI page for the command set.

Not sure whether your agent should use your inbox or its own? Read your email or its own?

What every agent gets

Its own address, like name@youreplyz.com, separate from your personal email.
Wait for email in one call, with a timeout to retry.
Codes and links pulled out of the email for it.
Your approval before anything is sent, and no cold or bulk email.
EU hosting on every plan.

Frequently asked questions

Which AI agents can use Replyz?

Claude Code, OpenClaw and Hermes through the skill and CLI, and Claude, Codex, Cursor and other MCP clients that can send an API key header through the MCP server.

Should my agent use the skill, the CLI or MCP?

Agents that run on your machine and can run commands (Claude Code, OpenClaw, Hermes, Codex) work best with the skill and CLI. Agents in a browser or app (Claude on the web, Desktop, Cowork) use the MCP server.

Does ChatGPT work with Replyz?

Not yet. ChatGPT connectors need OAuth, which Replyz plans to add after launch.

Can one person run several agents on Replyz?

Yes. Each agent or task can have its own inbox under the same account.