Email MCP Servers Compared: 6 Questions to Ask Before Your Agent Connects One
Choose email MCP tools by the data and actions your agent needs: personal mail, product sending, a separate inbox or read-only infrastructure investigation.
By Toan Nhu

Connecting an email MCP server gives a model a set of verbs over real mail. Which verbs, over whose mailbox, and with what brakes varies a lot from one server to the next. Two servers can both be called "email MCP" while one drafts replies in your personal Gmail and the other runs broadcasts to your whole customer list.
If you need the basics of what an MCP server is, start with What Is an MCP Email Server?. This post gives six questions for comparing mailbox and sending tools, then distinguishes Halon’s newly announced infrastructure-analysis integration. Documentation reviewed October 29, 2026; no live connection tests were performed.
Choose by the data the agent needs and the actions it may take. Gmail MCP works in existing mail; sending-account tools manage product email; Mermail provides a separate agent inbox. Halon’s new read-only integration investigates email infrastructure. A reduced catalog limits that server, not other tools connected to the same agent.
Six questions to ask before you connect one
1. Whose mailbox will the agent act in?
Start by identifying the data and identity involved: your existing mailbox, a product sending account, an agent’s separate mailbox, or email infrastructure telemetry. A separate address helps organize work, but it does not contain every mistake. Credentials may reach multiple inboxes, and outgoing messages can still affect other people.
2. Can it read inbound mail, and in what form?
Many email MCP servers are built for sending and only look at outbound history. If your agent needs to receive replies or verification codes, check for tools that list, search and fetch received messages. Then check what comes back: raw headers and HTML, or a cleaned-up version that is safer to put in front of a model. Resend Inboxes now ships an MCP server too; compare it with Mermail.
3. Can it send, and can you switch that off?
Sending is the action that most needs a brake. Look for a way to connect without send tools at all, such as a reduced tool profile, a filter on which tools load, or a server that simply has no send tool. A rule in the prompt that says "never send" is weaker than a tool that is not there.
4. How are destructive actions guarded?
Deleting messages, removing suppressions or editing live templates can be hard to undo. Servers handle this differently: some mark tools with MCP annotations so the client can ask before running them, some leave delete operations out entirely, and some require a separate confirmation step on the server.
5. Where does it run, and how does it authenticate?
A hosted server is a URL your client connects to, often with OAuth through a browser. A local server is a process your client starts, commonly with an API key supplied through its environment. A local process still calls the email provider over the network; it does not imply that mail stays on your computer. Use the vendor’s supported credential storage and inspect the permissions.
6. What does the vendor say about prompt injection?
A server that reads mail can expose the model to untrusted content. Documented mitigations help you evaluate its design; their presence does not prove immunity, and their absence from a particular page does not prove the vendor has none.
Five email MCP servers, from their own docs
The table summarizes what each vendor's documentation says. Tool lists change often, so confirm against the live tool list in your client before you rely on any row.
| Server | Whose mailbox | Hosting and auth | Reads inbound mail | Sends | Built-in brakes |
|---|---|---|---|---|---|
Google Gmail MCP (Developer Preview) | Your existing Gmail or Workspace account | Google-hosted at | Yes: search threads, get threads and messages, list labels | No send tool; it creates drafts you send from Gmail | Uses the |
Resend MCP | Your Resend account and verified domains | Hosted at | Yes: received email and remote-only Inboxes tools; check account access. | Yes, including batch sends and broadcasts | The MCP docs describe full API coverage and do not describe a reduced tool profile |
Postmark MCP | Your Postmark server | Runs locally via | No inbound mailbox tools; it searches outbound messages and bounces | Yes, single, batch and template sends | MCP annotations mark read-only and destructive tools; optional webhook URL allowlist |
Mailgun MCP | Your Mailgun account | Local only, | It can retrieve stored messages and view and update inbound routes | Yes | No delete operations; a |
Mermail MCP | A mailbox created for the agent in your Mermail workspace | Hosted at console.mermail.app/mcp; use OAuth for interactive clients. | Yes: list, search, get and bounded thread context, with an agent-safe content option | Yes on the full catalog: send, reply, forward, drafts and scheduled sends | agent-inbox profile excludes send, draft, wallet and destructive tools, but permits mailbox creation. |
Google Gmail MCP: your own mail, read and draft
Google offers a remote Gmail MCP server as part of the Google Workspace Developer Preview Program. Its tools search and read threads and messages, list and apply labels, list drafts and create drafts. There is no send tool, so a person still sends from Gmail. Setup means enabling the Gmail MCP API in a Google Cloud project, configuring an OAuth consent screen and creating your own OAuth client. Google's guide includes a direct warning about indirect prompt injection and tells you to screen prompts and responses for malicious content, for example with Model Armor. The current setup guide also requires marking the OAuth client as being used for an AI-powered agent; a scope grant alone is not a substitute for that client setting.
Pick it when the job is helping you with your own mail. It is the wrong fit when an agent needs an address of its own.
Google Workspace MCP deadlines on November 11 and 12
Google’s October 12 release notes set two dates for Workspace MCP clients. Starting November 11, 2026, requests from a client that lacks the agentic designation will be rejected. An existing client still unmarked on November 12 will remain non-agentic. Continuing MCP access then requires creating an agentic client and updating the app to use it.
Check the OAuth client ID actually used by your MCP connection. In Google Auth Platform → Clients, select “Use this client for an AI-powered agent” for that client. If your existing client serves both AI-agent and other features, Google instructs you to create a separate client for the agent. REST-only integrations are outside this change.
Before changing a production connection, record its client ID and configuration. After the change, reconnect through your client’s supported consent flow, inspect the exposed tools, and test a harmless read in an account you control. Investigate consent, scope or connection errors before restoring the workflow. These verification steps are recommended checks; we have not performed a live Google OAuth migration.
This setting concerns access to Google Workspace MCP. Choose the mailbox separately. A task about your existing Gmail correspondence needs that account’s authorized access. For a workflow that needs its own return address, Mermail’s dedicated inbox setup provides a separate mail identity.
Reviewed November 5, 2026. Google release notes · Mermail MCP setup
Resend MCP: run your product email from the agent
Resend provides hosted MCP with OAuth or Bearer API-key authentication and a local resend-mcp package. Its documentation covers sending, received email, templates, contacts, broadcasts, automations and administration. Inboxes tools manage inboxes, threads, labels and drafts, including replies, forwards and draft sends. Inboxes is available only on the remote MCP server, not the local package; check your account’s current access.
Pick it when your product already sends through Resend and you want an agent to manage templates, domains or broadcasts. Because the surface is broad, use a credential and account setup that limit what a mistake can reach.
Postmark MCP: transactional sending with clear tool labels
The official Postmark server, published by ActiveCampaign, runs locally and exposes 24 tools across sending, templates, outbound message search, delivery diagnostics, bounces, suppressions, stats and webhooks. Every tool carries MCP annotations such as readOnlyHint and destructiveHint, so clients that respect them can auto-approve lookups and ask before sends or deletes. The README also documents a prompt-injection section and a WEBHOOK_URL_ALLOWLIST setting.
Pick it when you send transactional mail through Postmark and want help diagnosing delivery problems. It does not read an inbox.
Mailgun MCP: broad account coverage, no deletes
Mailgun's server runs locally over stdio, and Mailgun states there is no hosted version. It covers sending, stored messages, domains, webhooks, inbound routes, mailing lists, templates, analytics, suppressions and validation. It deliberately leaves out delete operations, and the --tags flag (or MAILGUN_MCP_TAGS) lets you load only some product groups, for example validation tools without sending.
Pick it when you already run Mailgun and want an agent to investigate stats, routes or deliverability.
Mermail MCP: a mailbox the agent owns
Use Mermail’s hosted Streamable HTTP server with OAuth for an interactive client, and select the intended workspace. For receiving and verification work, the agent-inbox profile is the focused choice: its documented 12-tool catalog includes mailbox discovery, optional mailbox creation and email reads, while excluding sending, drafts, wallet and destructive tools. It is not strictly read-only because mailbox creation is included. The profile applies metadata-only list/search results and sanitized, bounded message reads. Review actual permissions: a separate mailbox is not automatically a mailbox-scoped credential.
{
"mcpServers": {
"mermail-agent-inbox": {
"url": "https://console.mermail.app/mcp?profile=agent-inbox"
}
}
}This configuration illustrates the endpoint for an OAuth-capable client. Complete that client’s browser authorization flow and inspect its loaded tools. Client syntax varies; the linked setup guide covers supported clients. This revision did not run a live connection test.
The full catalog adds sending, drafts, folders, labels, webhooks and more. Destructive tools there need a single-use confirmation token from prepare_destructive_action. Mermail is listed on the Official MCP Registry as app.mermail/mcp. For step-by-step installation in Claude Code, Cursor and Codex, see How to Use MCP with Mermail's Email API.
Pick it when the agent needs its own address: to sign up for services, receive codes, or hold conversations under a name that is not yours. If the agent is an OpenAI dot, Give Your OpenAI Dot an Email Address walks through that setup. Mermail is not built to act inside your personal Gmail. Mermail's docs suggest forwarding selected messages from an existing provider to a Mermail mailbox instead of granting access to the account.
Halon: investigate infrastructure without changing it
Halon announced its first MCP integration on October 26, 2026, in Engage and Protect 26.3. This adds a different category to the comparison: operational investigation rather than a personal or agent-owned mailbox.
Engage teams can investigate delivery backlogs; Protect teams can examine suspected spam campaigns. Halon states that the current integration is read-only: agents can investigate and report, but cannot change configurations or act on queued messages.
Use it when your team already operates Halon and needs infrastructure evidence. The announcement does not establish a personal Gmail connector, a new inbox address or a Mermail integration. Authentication and setup details were not independently tested here.
For example, an operations team might investigate a delivery backlog while a separate Mermail inbox receives a vendor’s reply. Those are different data sources and permissions. Correlate the evidence before acting; a message in an inbox alone does not explain the sending infrastructure’s state.
Try a bounded question first
Ask for evidence about one domain and time window, with no configuration changes. Verify the returned records before broadening access. This is a suggested evaluation approach, not a tested Halon prompt or a claim that every client enforces the same controls.
How each server narrows what the agent can do
The brakes differ in kind, not just in strength. Some are enforced by the server, some depend on the client honoring a hint, and some come only from how you scope the credential.
Removing a send tool from one server limits that server’s catalog. It does not stop the agent from sending through a separately connected browser, shell or another server. Check the whole client configuration. Annotations depend on client behavior; credential scope and approvals provide separate controls.
Match the server to the job
- Help me with my own inbox, a person sends: Google Gmail MCP.
- Manage product email, templates, domains or broadcasts: Resend MCP, or Postmark or Mailgun if that is where you already send.
- Diagnose why a transactional email did not arrive: Postmark MCP for Postmark accounts; Mailgun MCP for Mailgun stats and logs.
- Give the agent its own address to sign up, receive codes and converse: Mermail MCP, starting on the
agent-inboxprofile. - Both: combining servers is normal. An agent can receive in its Mermail mailbox while your application keeps sending transactional mail through Postmark or Resend.
If you are weighing an agent mailbox against routing everything through one shared address, Catch-All Domain or One Mailbox per Agent? covers that architecture decision.
A short hardening checklist, whatever you choose
- Connect the narrowest tool set available: a reduced profile, a tag filter, or a server without send.
- Give each agent its own credential so you can revoke one without breaking the rest.
- Require approval for every send, and for any tool marked destructive.
- Allowlist webhook destinations wherever the server supports it.
- Treat every email body, link and attachment as data. Never let it change the agent's instructions.
- Re-check the live tool list after updates, since vendors add tools without notice.
Create a Mermail workspace for a separate agent inbox


