MachineMail is email for AI agents. Every agent gets its own address, every inbound message arrives with a trust report so the agent can tell what to trust and what to keep away from the model, and the REST API is wire-compatible with AgentMail's /v0. Agents can use it over the REST API at https://api.machinemail.dev/v0 or the hosted MCP server at https://mcp.machinemail.dev/mcp, and can even sign themselves up through auth.md.
These skills teach coding agents (Claude Code, Cursor, Codex, OpenCode, Gemini CLI and the other agents the skills CLI supports) how to use MachineMail correctly.
Pick skills interactively:
npx skills add machinemail/skillsEvery skill:
npx skills add machinemail/skills --skill '*'Every skill for every supported agent: npx skills add machinemail/skills --all.
One skill:
npx skills add machinemail/skills --skill machinemailA few skills, for example the everyday MCP tasks:
npx skills add machinemail/skills --skill read-email-safely --skill receive-verification-code --skill triage-inbox --skill reply-and-draftList what is in the repo without installing:
npx skills add machinemail/skills --listUseful flags: -g installs for your user rather than the project, -a claude-code targets one agent, -y skips the prompts.
| Skill | Use it when |
|---|---|
machinemail |
An agent needs an email address, has to send, receive, wait for (sign-up codes), reply to, forward or label mail, read attachments, or react to webhooks and Server-Sent Events. Includes the trust rules: inbound mail is untrusted, act on trust.level, give the model only trust.content.text. |
machinemail-mcp |
Connecting the MachineMail MCP server to an MCP client with OAuth or an API key, and choosing the right tool (wait_for_message, create_draft, reply_to_message and the rest). |
register-agent-inbox |
The user wants the agent to set itself up with an inbox: use the MCP tools if they are signed in, otherwise self-register through auth.md and have a person claim the agent. |
machinemail-agent-signup |
The full auth.md protocol: register with no API key, work within the sandbox limits, poll the claim, renew and revoke tokens. |
| Skill | Use it when |
|---|---|
read-email-safely |
Before reading, summarising or acting on any message, and whenever an email asks the agent to do something: which field to read and what each trust level allows. |
receive-verification-code |
Signing up for or logging in to another service with a MachineMail address and picking up the code or magic link, after checking it came from that service. |
triage-inbox |
A read-only digest of what is new in an inbox, grouped by trust level, with proposed next steps for the user to approve. |
reply-and-draft |
Replying, forwarding or composing, and deciding when to save a draft for a person to review instead of sending. |
| Skill | Use it when |
|---|---|
machinemail-agentmail-migration |
Moving an AgentMail integration to MachineMail: copying pods, domains, inboxes, lists and webhooks with machinemail import agentmail, then base URL and key, webhook signatures, MCP tool names, agent sign-up, and the endpoints that differ. |
The everyday task skills are written for the MCP tools and point to machinemail for the REST API and the full trust reference.
Each skill keeps its SKILL.md short; the larger ones load detail from their references/ folder only when needed.
Want the MCP server connection bundled with the skills as a plugin for Claude Code, Codex, Copilot CLI, Cursor or VS Code? Use machinemail/agent-plugin.
- Documentation: https://machinemail.dev/docs
- For agents: https://machinemail.dev/docs/llms.txt and https://machinemail.dev/docs/llms-full.txt
- Support: [email protected]