Skip to main content
Glama
WeeCi

Mektup-mcp

README.md
# Mektup MCP Server

[Mektup](https://usemektup.com) 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](https://usemektup.com/docs/api.md). 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):

- **Remote (Streamable HTTP)** — a server we host at `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.
- **Local (stdio)** — run `server.js` yourself with your key in an env var. For MCP clients that only support launching a local process (Claude Desktop config, etc.).

---

## Remote server (recommended for platform integrations)

**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 <token>` header:

### OAuth (recommended for platform integrations)

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.

### API key (simplest for a single account, scripts, or a client without OAuth support)

`Authorization: Bearer mek_live_...` — create one at [app.usemektup.com](https://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:**
```json
{
  "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.

---

## Local (stdio) setup

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](https://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:**
```bash
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`):

```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.

---

## How tools respond

Every tool returns its result as JSON text on success. On failure, it returns `isError: true` with `Error: <message>` — the message is the same one the underlying REST endpoint returned (see the [API reference](https://usemektup.com/docs/api.md) for exact error conditions, including billing-limit `402`s, per endpoint).

---

## Tools

### Account

| 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. |

### API keys

| 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. |

### Domains

| 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. |

### Mailboxes

| 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. |

### Webhooks

| 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. |

### Forwarding

| 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. |

### Identity

| 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. |

### Vacation / auto-reply

| 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. |

### Folders

| 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). |

### Contacts

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. |

### Drafts

| 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. |

### Sending

| 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"] } }
```

### Messages & threads

| 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 already has filename/type/size. |

### Account-wide

| Tool | Input | Description |
|---|---|---|
| `get_unread_counts` | — | Unread Inbox count for every domain/mailbox at once. |

---

## See also

- [Full REST API reference](https://usemektup.com/docs/api.md) — everything this server wraps
- [OpenAPI 3.1 spec](https://usemektup.com/docs/openapi.yaml) — machine-readable version of the same API
- [usemektup.com](https://usemektup.com) — sign up, dashboard, pricing

## License

MIT

TDQS

A3.6/5.0

Scored across 45 tools

Disambiguation5/5

Every tool targets a distinct resource and action—account, domains, mailboxes, forwards, identity, vacation, folders, contacts, drafts, and messages are clearly separated. The only potential confusion is between list_messages/get_thread/get_message, but the descriptions explain the hierarchy. No two tools appear to do the same thing.

Naming Consistency4/5

Tools overwhelmingly follow a verb_noun snake_case pattern (list_*, get_*, create_*, update_*, delete_*). Minor deviations like add_forward/remove_forward, reset_mailbox_password, verify_domain, send_email, and download_attachment are still verb-first and readable, but they break the strict CRUD verb consistency.

Tool Count2/5

At 45 tools, the server far exceeds the typical well-scoped range and crosses the 'too many' threshold. Each tool is individually justifiable given the breadth of subdomains (account, domains, mailboxes, webhooks, contacts, drafts, messages), but the overall surface is heavy for agents to navigate.

Completeness4/5

The tool set provides CRUD/lifecycle coverage for almost every resource: domains, mailboxes, webhooks, contacts, drafts, and messages. Minor gaps exist—no explicit send-draft action, no attachment handling in send_email, no folder rename or single-mailbox get—but agents can work around them.

Maintenance

ActivityMaintained
ResponsivenessNo issues