MachineMail
Email API that gives AI agents their own inboxes, with a trust report on every inbound message and a hosted MCP server.
At a Glance
For your first agent.
Engagement
Available On
Alternatives
Listed Nov 2026
About MachineMail
MachineMail is an email API for AI agents. It lets software that reads and writes its own mail create an inbox with one API call, send and receive from it, and get each inbound message back as structured JSON with an agent-safe body and a trust verdict. Agents can connect through a hosted MCP server, a REST API, or a CLI.
What It Is
MachineMail provides real email addresses for agents. Inboxes live on the shared machinemail.dev domain, or on a custom catch-all domain where any address becomes an inbox on first delivery. Ephemeral inboxes expire on their own, and a reaper deletes the inbox with its messages, raw MIME and attachments.
The /v0 REST API follows AgentMail's paths, field names and enums, so an existing client can switch by changing the base URL and API key.
How the Trust Layer Works
Every inbound message is checked before it is stored and gets a trust report and a level: verified, known, unknown, suspicious or quarantined. The checks combine SPF, DKIM and DMARC alignment, allow and block lists, prior correspondence, hidden text and zero-width character removal, link and attachment checks, and a classifier for prompt injection, phishing and credential requests. The agent reads the cleaned text and acts on the level, never on what the sender says. Quarantined threads are hidden from agents unless they list with include_quarantine, and a person can review the report and release or block them.
Agent Workflow and Delivery
An agent can add the MCP server, for example to Claude Code, then create inboxes, wait for messages, read, reply and check trust. The wait_for_message tool holds the call open until mail lands, for up to a minute, which suits sign-up verification codes. Agents can also self-register through auth.md, receive a sandbox inbox and access token, and show a six-digit code so a human can claim them into an account. Real-time delivery is available through signed webhooks that are retried for five days and replayable from the dashboard, plus Server-Sent Events and WebSockets.
Skills and Tooling
The MIT-licensed machinemail/skills repository teaches coding agents such as Claude Code, Cursor, Codex, OpenCode and Gemini CLI how to use MachineMail. Its skills cover reading email safely, receiving verification codes, triaging an inbox, replying and drafting, MCP setup, agent sign-up, and migrating from AgentMail. A separate agent plugin bundles the MCP connection for several clients.
Community Discussions
Be the first to start a conversation about MachineMail
Share your experience with MachineMail, ask questions, or help others learn from your insights.
Pricing
Free
For your first agent.
- 3,000 messages a month, up to 100 a day
- 1 GB storage
- 1 custom domain
- 1 seat
- Unlimited inboxes
Pro
For agents in production.
- 7,500 inbound and 2,500 outbound a month
- 10 GB storage
- 10 custom domains
- Unlimited seats
- Overage: Inbound $1.00 per 1,000 messages
- Overage: Outbound $0.50 per 1,000 recipients
- Overage: Storage $0.25 per GB per month
- Overage: Custom domain $1 per domain per month over the plan
- Email support
Team
For teams of agents.
- 45,000 inbound and 15,000 outbound a month
- 30 GB storage
- 50 custom domains
- Unlimited seats
- Overage: Inbound $1.00 per 1,000 messages
- Overage: Outbound $0.50 per 1,000 recipients
- Overage: Storage $0.25 per GB per month
- Overage: Custom domain $1 per domain per month over the plan
- Email support
Scale
For fleets of agents.
- 200,000 inbound and 65,000 outbound a month
- 150 GB storage
- 250 custom domains
- Unlimited seats
- Dedicated Slack support
- Overage: Inbound $1.00 per 1,000 messages
- Overage: Outbound $0.50 per 1,000 recipients
- Overage: Storage $0.25 per GB per month
- Overage: Custom domain $1 per domain per month over the plan
Capabilities
Key Features
- One-call inbox creation on a shared or catch-all custom domain
- Ephemeral inboxes that expire automatically
- Trust report and trust level on every inbound message
- Agent-safe text with hidden instructions and zero-width characters removed
- Quarantine of suspected prompt injection and phishing
- Hosted MCP server with wait_for_message
- Webhooks, Server-Sent Events and WebSockets
- auth.md agent self-registration with human claim
- AgentMail-compatible /v0 REST API
- CLI and OpenAPI document
- Agent skills for coding agents
