Vexa
Meeting bot and transcripts for Google Meet, Teams and Zoom. Live or after, speakers labelled.
作者提供的说明(英文原文)
Fully-managed email as MCP tools - register domains, real mailboxes, send and receive mail.
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
选择你使用的 AI 应用,然后复制设置。
Authorization标为机密的值是你账户的钥匙。千万不要分享给别人,也不要粘贴到聊天里。
你需要
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
这个地址需要在请求的 header 中放入密钥,而添加连接器的菜单无法设置。请改用 Claude Code、Cursor 或 VS Code。
在终端中运行这条命令
claude mcp add --transport http mektup https://mcp.usemektup.com/mcp --header 'Authorization: your-value'运行前请先填写空着的值。
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"mektup": {
"url": "https://mcp.usemektup.com/mcp",
"headers": {
"Authorization": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"mektup": {
"type": "http",
"url": "https://mcp.usemektup.com/mcp",
"headers": {
"Authorization": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
这个地址需要在请求的 header 中放入密钥,而添加连接器的菜单无法设置。请改用 Claude Code、Cursor 或 VS Code。
你的选择通过 Cookie 保存在这台设备上,无需账户。
Authorization 机密 必填 HTTP header这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
我们还没有读取这个服务器的工具列表;我们会轮流检查各个端点。
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月7日
注册名称已通过 GitHub 账户 WeeCi 验证,该账户拥有此仓库
我们还没有从 GitHub 读取这个仓库
我们还没有从 GitHub 读取这个仓库
尚未检查;我们会轮流检查各个端点
尚未检查;我们会轮流检查各个端点
尚未检查;我们会在检查端点时读取工具列表
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 1 项适用的基础检查中通过了 1 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
你需要填写:Authorization
它在作者的云端运行。通过互联网连接,无需安装,但你发送的内容会经过作者的服务器。
GitHub 仓库最后更新于–,最新版本(0.2.6)于3个月前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。它通过 URL 连接,所以接受远程 MCP 服务器的应用也可以使用,例如开启开发者模式的 ChatGPT。
同一分类中有:Vexa, emailmd, MCP WhatsApp Cloud
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
Meeting bot and transcripts for Google Meet, Teams and Zoom. Live or after, speakers labelled.
Render markdown into email-safe HTML, lint drafts for deliverability problems, and preview emails.
MCP server for WhatsApp Cloud API (Meta) — messages, media, and templates via Graph API
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
Mektup is a fully-managed, multi-domain email platform - Mektup hosts all the mail infrastructure, so there's nothing to self-host; you just sign up and use it. This MCP server lets an AI coding agent (Claude Code, Claude Desktop, Cursor, Lovable, Replit, Base44, or any MCP-compatible client) manage real email for a domain — register it, add the DNS records, create mailboxes, send and read mail, manage drafts/contacts/folders/forwarding/vacation replies — as native tool calls inside its own session, instead of the human hand-writing curl commands or pasting an API key into generated code.
Every tool call is a straight HTTP call to the real Mektup REST API. There's no separate logic to learn — if you understand the API, you understand the MCP server. Full coverage: every REST endpoint has a matching tool, verified with real read and write round-trips against the live production API (create → update → list → delete, confirmed at each step).
Two ways to run it, same tool set either way (lib/build-server.js defines the tools once, shared by both):
https://mcp.usemektup.com/mcp. Point any client that takes a "custom MCP server" URL at it directly, no install. This is what Lovable/Cursor/Replit/Base44-style platforms want.server.js yourself with your key in an env var. For MCP clients that only support launching a local process (Claude Desktop config, etc.).Endpoint: https://mcp.usemektup.com/mcp (Streamable HTTP — supports both the direct-JSON-response and SSE-streaming response modes of the spec).
Stateless: no session is kept between requests — every tool call is already a one-shot passthrough to the REST API, so there's no session state worth keeping.
Tenant isolation: identical to the REST API, because it *is* the REST API underneath — the server holds no account-specific credential itself, it just forwards whichever token the caller sent (API key or OAuth access token, see below) straight through to api.usemektup.com, which is the only place that actually verifies it. A token only ever sees its own account's data.
Two auth modes, same endpoint, both arrive as the same Authorization: Bearer header:
For a platform with real end users (Lovable, Cursor, Replit, Base44, ...) — the user clicks "connect," signs in with their existing Mektup account, approves, done. No token copy-paste, no dashboard visit.
Clerk (clerk.usemektup.com) is the OAuth 2.1 authorization server — this MCP server is only ever a resource server. Discovery is automatic for any spec-compliant OAuth-capable MCP client: it only needs the endpoint URL above and finds the rest itself via https://mcp.usemektup.com/.well-known/oauth-protected-resource/mcp (RFC 9728), which points at Clerk's own https://clerk.usemektup.com/.well-known/oauth-authorization-server (RFC 8414). From there the client registers itself via Dynamic Client Registration (no manual setup needed on your end) and runs a standard Authorization Code + PKCE flow, ending with a JWT access token used exactly like an API key.
Lovable: Connectors → All → Custom (MCP card) → Server Name Mektup, Server URL https://mcp.usemektup.com/mcp, Auth → OAuth (default when a server supports it) → Add & authorize.
Authorization: Bearer mek_live_... — create one at app.usemektup.com → API keys.
Adding it to a client that supports custom MCP connectors is normally 3 fields:
| Field | Value |
|---|---|
| Server URL | https://mcp.usemektup.com/mcp |
| Auth type | Bearer token / API key |
| Token | your mek_live_... key |
Cursor / Claude Desktop / any client that reads raw MCP JSON config:
{
"mcpServers": { "mektup": { "url": "https://mcp.usemektup.com/mcp", "headers": { "Authorization": "Bearer mek_live_..." } } } }
Replit / Base44 / other "connect a tool" flows: same three fields as the table above — server URL, Bearer auth, key.
Use this when a client can only launch a local process, not call a remote URL.
1. Get an API key. Sign in to the dashboard at app.usemektup.com, open API keys, and create one. Keys look like mek_live_... and are shown exactly once - copy it immediately.
2. Install dependencies:
git clone https://github.com/WeeCi/mektup-mcp.git
cd mektup-mcp npm install
3. Configure your MCP client to run server.js with the key as an environment variable. For Claude Desktop / Claude Code, add to your MCP config (e.g. claude_desktop_config.json):
{
"mcpServers": { "mektup": { "command": "node", "args": ["/absolute/path/to/mektup-mcp/server.js"], "env": { "MEKTUP_API_KEY": "mek_live_..." } } } }
MEKTUP_API_BASE_URL is optional and defaults to https://api.usemektup.com — no need to set it under normal use.
The server refuses to start without MEKTUP_API_KEY set.
Every tool returns its result as JSON text on success. On failure, it returns isError: true with Error: — the message is the same one the underlying REST endpoint returned (see the API reference for exact error conditions, including billing-limit 402s, per endpoint).
| Tool | Input | Description |
|---|---|---|
get_me | — | Get the authenticated account identity. Useful as an auth health check. |
get_usage | — | Current billing tier, its limits, and real usage against them. Check before a bulk operation. |
| Tool | Input | Description |
|---|---|---|
list_api_keys | — | List keys on this account (prefix and status only). |
create_api_key | — | Create a new key. The full key is returned exactly once - surface it to the user immediately so they can save it. |
revoke_api_key | id | Revoke a key immediately. Cannot be undone - confirm with the user first, especially if it might be the key this very session is using. |
| Tool | Input | Description |
|---|---|---|
create_domain | domain | Register a domain, get back the exact DNS records (MX/SPF/DMARC/DKIM) and a setup recommendation. Never touches DNS itself. |
list_domains | — | List every domain on this account. |
get_domain_records | domain | Re-fetch a registered domain's DNS records any time after creation. |
verify_domain | domain | Actively re-check live DNS and flip verified once it matches. Not automatic. |
delete_domain | domain | Delete a domain and everything under it. Destructive — confirm with the user first. |
| Tool | Input | Description |
|---|---|---|
create_mailbox | domain, localPart, password? | Create a mailbox with real IMAP/SMTP-AUTH credentials, usable in any mail client. |
list_mailboxes | domain | List mailboxes on a domain. |
reset_mailbox_password | domain, localPart, password? | Reset a mailbox's login password. Shown once. |
delete_mailbox | domain, localPart | Delete a mailbox. Confirm with the user first. |
| Tool | Input | Description |
|---|---|---|
get_mailbox_webhook | domain, localPart | Check a mailbox's configured webhook URL (never returns the signing secret). |
set_mailbox_webhook | domain, localPart, url, regenerateSecret? | Set/update the URL that fires (HMAC-signed) on every new inbound message - how an agent finds out about new mail without polling list_messages. Returns the signing secret once, on first setup or rotation. |
delete_mailbox_webhook | domain, localPart | Remove a mailbox's webhook. |
| Tool | Input | Description |
|---|---|---|
list_forwards | domain, localPart | List addresses that get a copy of incoming mail. |
add_forward | domain, localPart, forwardTo | Add a forwarding address. |
remove_forward | domain, localPart, id | Remove a forwarding address. |
| Tool | Input | Description |
|---|---|---|
get_identity | domain, localPart | Get display name and signature. |
set_identity | domain, localPart, displayName?, signatureText?, signatureHtml? | Set display name/signature, applied automatically to outgoing mail. |
| Tool | Input | Description |
|---|---|---|
get_vacation | domain, localPart | Get vacation auto-reply settings. |
set_vacation | domain, localPart, enabled, subject?, message? | Enable/configure auto-reply. message required when enabling. |
| Tool | Input | Description |
|---|---|---|
list_folders | domain, localPart | List custom folders. |
create_folder | domain, localPart, name | Create a folder. |
delete_folder | domain, localPart, id | Delete a folder (mail in it falls back to Inbox/Sent). |
Account-level, not per-mailbox.
| Tool | Input | Description |
|---|---|---|
list_contacts | — | List contacts. |
create_contact | name?, email | Add a contact. |
update_contact | id, name?, email? | Partially update — only send fields to change. |
delete_contact | id | Delete a contact. |
| Tool | Input | Description |
|---|---|---|
list_drafts | domain, localPart | List drafts (metadata only). |
get_draft | domain, localPart, id | Get a draft including its body. |
create_draft | domain, localPart, to?, subject?, text?, html? | Create a draft. |
update_draft | domain, localPart, id, to?, subject?, text?, html? | Partial update (autosave-friendly). |
delete_draft | domain, localPart, id | Delete a draft. |
| Tool | Input | Description |
|---|---|---|
send_email | from, to, subject, text?, html?, attachments?, draftId? | Send real mail. from must be a mailbox on this account, or any address on a domain this account has verified. Attachments are {filename, contentType?, contentBase64}, max 10MB decoded each. Pass draftId to delete a draft on successful send. |
Example:
send_email({ from: "hello@example.com", to: "you@gmail.com", subject: "It works", text: "Real mail, sent through Mektup." })
→ { "messageId": "<...@example.com>", "envelope": { "from": "hello@example.com", "to": ["you@gmail.com"] } }
| Tool | Input | Description |
|---|---|---|
list_messages | mailbox, limit?, direction?, trash?, folder?, q? | List messages (one row per thread). Pass direction to split Inbox/Sent — omitting it merges both. |
get_delivery_stats | mailbox, days? | Aggregate sent/deferred/bounced/unknown counts for a mailbox's outbound mail, sourced from Mektup's own Postfix delivery log — not a tracking pixel. |
get_thread | threadKey, mailbox, direction?, trash?, folder? | Every message in one thread, oldest first. |
get_message | id | Full message content. Marks it read as a side effect. `html` is attacker-controlled — never render it directly. |
update_message | id, read?, restore?, flagged?, folderId? | Mark read/unread, restore from trash, flag, or move to a folder — any combination in one call. |
delete_message | id | Two-stage delete: first call trashes, second call on an already-trashed message permanently deletes it. Confirm before a permanent delete. |
| download_attachment | id, index | Download one attachment, base64-encoded. Prefer only when the actual file content is needed — get_message's attachment list a
只显示了开头部分;其余内容请在仓库中阅读