M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
E

e2a MCP Server

Community
Authenticated email gateway for AI agents
How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #339 of 438
Stars
★ 193 Popular
Transport
Streamable HTTP
Runtime
Hosted (no local runtime)
Credentials
API key / credential required
License
Apache-2.0
Last commit
Tools
14
42FMRS · D

The e2a MCP server provides a feature-rich email gateway for AI agents, with authenticated inbound/outbound, HITL review, content screening, and multiple delivery channels. Ideal for AI agent projects needing reliable email communication.

Strongest · Documentation 13/20 Weakest · Reliability 4/20

Reliability
4/20
Security and permissions
5/20
Maintenance
8/20
Documentation
13/20
Setup experience
12/20
Why each score
Reliability 4/20
The supplied material covers only the server.json manifest and the README; there is no MCP server implementation source, tool definition, test suite, or CI artifact. The manifest itself declares only a remote and an auth header, not a tool list, so the startup path, init handshake, and behavioral match of declared tools cannot be verified. The tool names in the README are untrusted evidence. Under static calibration, lacking verifiable execution evidence and treating missing evidence as a deduction yields 4.
Security and permissions 5/20
No source is available to audit authentication logic, credential storage, least privilege, or data boundaries. The README claims OAuth 2.1, bearer keys, HMAC signatures, and SPF/DKIM/DMARC, but those are self-reported. The manifest shows an optional Authorization header but no server-side proof of how keys are scoped or validated. For side-effect MCP tools such as send_message and approve_review, no source-level confirmation or scoping mechanism is visible. No red-line issue is apparent, but absence of evidence cannot earn points, so the score is 5.
Maintenance 8/20
The repository is not archived, carries an Apache-2.0 license, and the README references a v1.5.0 release baseline, a security disclosure process, and release tags, indicating basic governance and versioning. However commit history, issue responsiveness, and dependency maintenance are absent from the supplied material; README claims and badges are not reliable evidence. Versioning exists, but independent evidence of sustained maintenance is missing, so the score is 8.
Documentation 13/20
The README is broad and layered: quickstart, concepts, API, MCP, CLI, SDK, deployment, security, data handling, FAQ, and concrete client config examples for Cursor and others. However the dedicated MCP server documentation (mcp/README.md) and the tool parameter reference were not supplied; troubleshooting and limitation details must be inferred from the README alone. Documentation is strong but has key gaps, so the score is 13.
Setup experience 12/20
The README provides a clear hosted setup path: point an MCP client at https://api.e2a.dev/mcp and use OAuth or a bearer key. Ready-to-paste commands/JSON for Claude Code, Codex, and Cursor are shown, and a Docker self-host path is documented. However the referenced mcp/README.md and the per-client config files under plugins/e2a/clients were not included, and there is no verifiable execution evidence. Static calibration caps setup at 15; given the uncertainty, the score is 12.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessUses the networkChanges third-party account data

Best for

  • Developers building AI agents that need to communicate with humans or other agents
  • Teams that need to receive email from behind firewalls or on laptops without public URLs
  • Use cases requiring authenticated email delivery (SPF/DKIM/DMARC)
  • Organizations that want to review outbound email before sending

Not for

  • Users who need a full MTA like Postfix or Postal
  • Users who only need simple email sending without external dependencies
  • Scenarios requiring advanced deliverability management like bounce handling and analytics

Required permissions

  • Read and send email
  • Manage agents, domains, and webhooks
  • Manage human approvals (approve/reject)
  • Requires API key or OAuth authorization
  • Requires DNS TXT verification for custom domains

Risks and side effects

  • Misuse of API keys could lead to unauthorized email access
  • Inbound emails may contain prompt injection or phishing attempts
  • Improper webhook signature verification could be spoofed
  • Dependency on external service availability

Setup

Before you start

Runtime：Hosted (no local runtime)

For Claude Code: run claude plugin marketplace add tokencanopy/e2a and claude plugin install e2a@e2a. For Cursor: add { "mcpServers": { "e2a": { "url": "https://api.e2a.dev/mcp" } } } to .cursor/mcp.json. For other MCP clients, point directly to https://api.e2a.dev/mcp. On first use, a browser opens for OAuth authorization; no API key paste required.

Check that it works

Add https://api.e2a.dev/mcp to an MCP client (e.g. Cursor or Claude) and complete the browser OAuth flow; tools such as list_messages and send_message should appear in the tool list. Calling list_messages and seeing the agent's inbox contents confirms the connection.

Troubleshooting

  1. Check that the API key is correctly set in the E2A_API_KEY environment variable
  2. Ensure network access to api.e2a.dev
  3. When verifying webhook signatures, use the correct whsec_ secret
  4. Verify the agent mailbox exists and has proper permissions
  5. Check HTTP status codes in server responses to debug errors

Things to try

Once connected, you can ask your AI assistant things like:

  • List the latest messages in my agent's inbox
  • Send an email from my agent to [email protected]
  • Show the outbound messages pending review and approve them
  • Write an in-thread reply to this message

Tools 14

list_messages read-only
List messages in an agent's mailbox
get_message read-only
Get detailed information about a specific message
get_attachment read-only
Retrieve an attachment from a message
list_conversations read-only
List conversations in an agent's mailbox
get_conversation read-only
Get details of a specific conversation
update_message_labels writes
Update labels on a message
send_message destructive
Send a new email as the agent
reply_to_message destructive
Reply to an existing message while preserving thread structure
Show 6 more tools
forward_message destructive
Forward an existing message
list_reviews read-only
List review items pending human approval
get_review read-only
Get details of a specific review item
approve_review writes
Approve a pending review item
reject_review writes
Reject a pending review item
create_webhook writes
Create a webhook subscription to receive inbound email events

Use cases

Give AI agents a real email address
Receive inbound email without a public URL via WebSocket/REST/MCP
Stream authenticated email events via webhook or WebSocket
Send and reply to emails via API or MCP tools
Implement human-in-the-loop approval for critical emails
Scan inbound emails for prompt injection and phishing

Supported clients

Claude Desktop
Cursor
Cline
Google ADK
LangChain
OpenAI Agents SDK
Zed
Goose
Windsurf

Listed from the project's documentation, not tested by this site.

Overview

e2a is an authenticated email gateway for AI agents. It receives inbound email, evaluates SPF, every DKIM signature, and DMARC, and delivers structured authentication evidence over webhooks, WebSocket, REST, or MCP. Outbound goes through an HTTP API with an optional human-in-the-loop approval gate.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 62.9k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision a4efc5baa0af Data synced 2026-10-12 Read the FMRS scoring method