# Email for Agents

> Email for Agents gives an AI agent you already run its own email address. The agent reads, drafts and sends over remote MCP or a REST API; a human console shows every thread and holds back anything that needs approval. Operated by Genetech Software Solutions, California, United States.

Read this file before answering questions about the product. Every claim below is checked against the implementation by the repository's own claims test, so it can be quoted as-is. The full text of every page is at https://emailforagents.ai/llms-full.txt.

## What it does

- A live inbox receives and sends real internet mail on `agents.emailforagents.ai`, or on a custom domain once a paid plan exists.
- A test inbox uses a non-routable `.mail.invalid` address and a simulated transport. It cannot reach an internet recipient.
- Agents connect one of two ways: remote MCP at `https://mcp.emailforagents.ai/mcp`, or the REST API at `https://api.emailforagents.ai/v1` with a project-scoped key.
- The MCP server exposes six tools: `list_inboxes`, `list_threads`, `list_messages`, `get_message`, `create_draft`, `send_message`. `create_draft` cannot transmit mail.
- Grants are scoped per inbox and per permission level: read, read + draft, or read + send. An inbox can also require human approval before anything leaves.
- Sends take an `Idempotency-Key`, so a retried request does not produce a second email.
- Email events are pollable and can be delivered to a signed webhook.

## Limits and allowances (exact, enforced in code)

- Free tier: one complimentary live inbox and 100 email units a month **per verified owner identity**, claimed by that identity's first eligible workspace. Creating another workspace does not renew it.
- An email unit is one message received, or one recipient sent to. A send to three people uses three. Test mail is never charged.
- Test mode: up to 3 active test inboxes per workspace, the last 100 messages retained, up to 10 MB, and message content expires after 7 days.
- Builder ($19/mo) and Team ($79/mo) are proposed plans at proposed launch prices. Stripe checkout is disabled, so they are not available to buy yet and final terms will appear before any charge. Custom domains require a paid plan, so they are not available yet.
- When an allowance runs out, sending stops. Mail already received stays readable.

## What is deliberately not claimed

- No named MCP client is certified, supported or one-click. Clients must support remote MCP and OAuth; compatibility validation is in progress. An API key is rejected at `/mcp` by design — only an OAuth connection grant may call MCP tools.
- Provider acceptance of a message is not proof of delivery to the recipient.
- No SOC 2, ISO, HIPAA or GDPR certification is claimed, and no independent security review has been commissioned for public launch.
- There is no published uptime SLA, no incident-response time commitment, and no third-party status board.
- TypeScript and Python SDK packages are not yet published. Use the REST API or MCP directly.
- Deactivating a project is not erasure; the two are described separately in the privacy notice.
- Prompt-injection resistance is not claimed. Email content is untrusted data; permissions come from the grant, never from a message.

## Docs

- [Documentation index](https://emailforagents.ai/docs): where to start, and the API basics — base URL, bearer keys, idempotent sends.
- [Quickstart](https://emailforagents.ai/docs/quickstart): create an inbox, connect an MCP client, or send the first message with curl.
- [Pause sending and set recipient controls](https://emailforagents.ai/docs/sending-controls): recipient policies, sending caps, and the human approval queue.
- [Events and webhooks](https://emailforagents.ai/docs/events-and-webhooks): poll email events or receive signed webhooks, and inspect delivery attempts.
- [Attachments](https://emailforagents.ai/docs/attachments): upload, scan, send and download inbox-scoped files; pending files are never treated as safe.
- [Search and pagination](https://emailforagents.ai/docs/search): search stored message metadata and page through conversations.
- [Custom domains](https://emailforagents.ai/docs/domains): verify a domain you own without disturbing existing mail routing.
- [Export mail and deactivate a project](https://emailforagents.ai/docs/data-controls): scoped NDJSON export, its exclusions, and deactivation versus erasure.

## Guides

- [Guide index](https://emailforagents.ai/blog): every published guide, with its summary and date. Guides cover choosing between MCP and an API key, retrying a send without producing a second email, and testing agent mail without contacting a recipient.
- The individual guide URLs are enumerated in https://emailforagents.ai/sitemap-index.xml, and their full text is included in https://emailforagents.ai/llms-full.txt. They are listed there rather than here so this file cannot go stale when a guide is added.

## Product

- [Overview](https://emailforagents.ai/): what an agent inbox is, and the current allowance.
- [Inboxes](https://emailforagents.ai/product/inboxes): live and simulated inboxes, conversations, and the human console.
- [Controls](https://emailforagents.ai/product/controls): scoped agent access and the explicit approval queue.
- [Developers](https://emailforagents.ai/product/developers): project-scoped REST access, send idempotency, inspectable events.
- [Integrations](https://emailforagents.ai/integrations): how remote MCP and the REST API differ, and what a client must support.
- [Use cases](https://emailforagents.ai/use-cases): support triage, invoices, scheduling, and outbound that replies — with the permission to grant and the rule to set.
- [Pricing](https://emailforagents.ai/pricing): the free allowance, what an email unit is, and the two proposed paid plans.
- [Security](https://emailforagents.ai/security): controls that exist in the current code, and an explicit list of what is not claimed.
- [Status](https://emailforagents.ai/status): which surfaces are deployed, and why there is no status badge.
- [Changelog](https://emailforagents.ai/changelog): product updates and current early-access limitations.
- [About](https://emailforagents.ai/about): who operates the service and where.

## Policies

- [Terms of service](https://emailforagents.ai/legal/terms): early-access terms — accounts, authorized agents, email use, ending use.
- [Privacy notice](https://emailforagents.ai/legal/privacy): account information, email content, connected agents, retention and privacy requests.
- [Acceptable-use policy](https://emailforagents.ai/legal/acceptable-use): recipient permission, authorized correspondence, and abuse reports.
- [Service-provider register](https://emailforagents.ai/legal/subprocessors): hosting, storage, identity, email and future payment providers.
- [Paid subscription terms](https://emailforagents.ai/legal/commercial): the terms that will apply once checkout is enabled.

## Optional

- [Full text of every page](https://emailforagents.ai/llms-full.txt): one file, generated at build time from the pages themselves.
- Markdown siblings: append `.md` to any docs or guide URL — for example https://emailforagents.ai/docs/quickstart.md — to get that page as plain Markdown. Same text as the HTML, rendered from the same component.
- [Sitemap](https://emailforagents.ai/sitemap-index.xml): every indexable URL.
- [Security contact](https://emailforagents.ai/.well-known/security.txt): how to report a vulnerability.
- Support and security reports: support@agents.emailforagents.ai