EngageLab Agent Email MCP Server
Enables AI agents to send, receive, and reply to email through dedicated thread-aware mailboxes, with sandbox testing and compatibility across MCP clients.
README
EngageLab Agent Email MCP Server
English | 简体中文
Send, receive, and reply to email from AI agents — through the Model Context Protocol.
Each agent gets its own mailbox: a real address that sends and receives, with conversations grouped into threads so your agent has context. Works with Claude Desktop, Claude Code, Cursor, Cowork, and any MCP-compatible client.
{
"mcpServers": {
"engagelab-email": {
"command": "npx",
"args": ["-y", "@engagelabemail/mcp"],
"env": { "ENGAGELAB_EMAIL_SECRET_KEY": "sk_sg_your_key" }
}
}
}
Why
Most email APIs are send-only. Agents that need to have a conversation over email — support, scheduling, approvals — need an inbox.
- Send & receive — the agent owns a dedicated mailbox, not a bolted-on parse webhook
- Thread-aware — replies group into conversations, so the agent sees full history
- Sandbox mode — test agent flows without emailing real humans
- One env var — no OAuth dance for local use; region inferred from the key
Tools
| Tool | What it does | |
|---|---|---|
| Discover | list_mailboxes |
List mailboxes; call this first to get IDs |
| Send | send_email |
New email (text/html, cc/bcc, attachments, sandbox) |
reply_email |
Reply to a message — recipients & thread inferred | |
| Receive | list_inbound_messages |
Inbound mail, filter by mailbox/keyword |
get_message |
Full message details by UID | |
check_new_messages |
One-shot poll for new mail | |
| Threads | list_threads / get_thread / list_thread_messages |
Browse conversations |
Quickstart
1. Get a Secret Key. Create an EngageLab account and generate a key from the console (format sk_sg_xxx — the prefix selects the region). Or use the CLI: engagelab-email-cli login creates one via browser.
2. Create a mailbox. In the console, create a mailbox for your agent (shared subdomain is instant; custom domains need DNS verification). Programmatic mailbox creation is on the roadmap — see Troubleshooting if list_mailboxes returns empty.
3. Register the server.
Claude Desktop — add the JSON block above to claude_desktop_config.json.
Claude Code:
claude mcp add engagelab-email \
-e ENGAGELAB_EMAIL_SECRET_KEY=sk_sg_your_key \
-- npx -y @engagelabemail/mcp
Then ask your agent:
List my mailboxes, then send an email from the first one to bob@example.com saying the invoice is approved.
A typical agent loop for a support bot:
check_new_messages() → [messageUids]
get_message(uid) → body, attachments
list_thread_messages(threadId) → full conversation context
reply_email(uid, { text: ... }) → response lands in the same thread
Configuration
| Environment variable | Required | Description |
|---|---|---|
ENGAGELAB_EMAIL_SECRET_KEY |
Yes | Secret Key starting with sk_ |
ENGAGELAB_EMAIL_BASE_URL |
No | Override the API base URL. Inferred from the key region: sg → Singapore, tr → Türkiye |
Limits: up to 10 attachments / 10MB total per message (base64-encoded in the tool schema).
Troubleshooting
list_mailboxesreturns an empty list — your account has no mailbox yet. Create one in the EngageLab console (see Quickstart step 2). Agents cannot self-provision mailboxes yet.- 401 /
code 100101— the Secret Key is wrong or revoked. CheckENGAGELAB_EMAIL_SECRET_KEYand that the key's region prefix (sk_sg_/sk_tr_) matches the endpoint you use. - Could not determine API base URL — the key's region prefix isn't recognized. Set
ENGAGELAB_EMAIL_BASE_URLexplicitly. - Reply shows up as a new conversation — use
reply_email(notsend_email) so In-Reply-To headers and threading stay intact.
Roadmap
- [ ] Hosted MCP (OAuth) for zero-config remote use
- [ ] Programmatic mailbox creation via Secret Key
- [ ] Webhook push (no polling) for
check_new_messages - [ ] URL-based attachment references (no base64 in context)
Run from source
git clone https://github.com/Metaverse-Cloud/engagelab-email-mcp
cd engagelab-email-mcp
npm install && npm run build
ENGAGELAB_EMAIL_SECRET_KEY=sk_sg_xxx node dist/index.cjs
Related
- EngageLab Agent Email CLI — same mailboxes from the terminal,
--jsonoutput for agents - EngageLab Node SDK / Python SDK — send email from code
License
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。