mailpit-mcp-server
Enables AI assistants to read, search, and manage emails from Mailpit, an email and SMTP testing tool, during development and testing.
README
mailpit-mcp-server
A Model Context Protocol server for Mailpit, the email & SMTP testing tool. It lets AI assistants (Claude Code, Claude Desktop, Cursor, VS Code, …) read, search and manage the emails your application sends during development and testing.
- Streamable HTTP transport — run it as a network service next to Mailpit (Docker Compose friendly), or
- stdio transport — let your MCP client spawn it locally
- Bearer-token auth for the HTTP endpoint, so it can safely sit behind a tunnel/reverse proxy
- Single small dependency surface: the official
@modelcontextprotocol/sdk, Express and Zod
Tools
| Tool | Description |
|---|---|
list_messages |
List mailbox messages, newest first (paginated) |
search_messages |
Search with Mailpit query syntax (from:, subject:, is:unread, tag: …) |
get_message |
Full message by ID or latest — headers, text/HTML bodies, attachment metadata |
get_message_headers |
Raw headers of a message |
get_message_source |
Full raw RFC-822 source (headers + MIME parts) |
get_attachment |
Download an attachment — images returned as images, text as text, other as base64 |
check_html |
HTML compatibility analysis across email clients |
get_message_links |
Extract all URLs without requesting them — safe retrieval of confirmation/reset/unsubscribe links for e2e flows |
check_links |
Verify links by requesting them (real GETs — may trigger one-click actions; auth-protected links can return 302/401/403 without being broken) |
send_message |
Compose and inject a test email via the Mailpit API (captured, not delivered) |
wait_for_message |
Poll until a new (optionally matching) message arrives — great for e2e flows |
set_read_status |
Mark messages read/unread, by ID or all |
delete_messages |
Delete messages by ID, or all messages |
get_mailbox_info |
Mailpit version, message/unread counts, database size |
Quick start (Docker Compose)
services:
mailpit:
image: axllent/mailpit:latest
ports:
- "8025:8025" # web UI
- "1025:1025" # SMTP
healthcheck:
test: ["CMD", "/mailpit", "readyz"]
interval: 30s
start_period: 10s
mailpit-mcp:
image: ghcr.io/cedjey/mailpit-mcp-server:1
ports:
- "3000:3000" # MCP endpoint: http://localhost:3000/mcp
environment:
MAILPIT_URL: http://mailpit:8025
# Recommended: require "Authorization: Bearer <token>" on /mcp
# MCP_AUTH_TOKEN: your-secret-token
depends_on:
mailpit:
condition: service_healthy
Or standalone against an existing Mailpit:
docker run -d -p 3000:3000 \
-e MAILPIT_URL=http://your-mailpit-host:8025 \
-e MCP_AUTH_TOKEN=your-secret-token \
ghcr.io/cedjey/mailpit-mcp-server:1
Connecting clients
Claude Code:
claude mcp add --transport http mailpit http://localhost:3000/mcp \
--header "Authorization: Bearer your-secret-token"
.mcp.json / Cursor / VS Code:
{
"mcpServers": {
"mailpit": {
"type": "http",
"url": "http://localhost:3000/mcp",
"headers": { "Authorization": "Bearer ${MAILPIT_MCP_TOKEN}" }
}
}
}
Claude Desktop (stdio):
{
"mcpServers": {
"mailpit": {
"command": "docker",
"args": ["run", "-i", "--rm",
"-e", "MCP_TRANSPORT=stdio",
"-e", "MAILPIT_URL=http://host.docker.internal:8025",
"ghcr.io/cedjey/mailpit-mcp-server:1"]
}
}
}
Or without Docker (Node 20+): clone the repo, npm ci, and use "command": "node", "args": ["/path/to/server.js"] with MCP_TRANSPORT=stdio.
Built-in agent context
The server ships the context an AI agent needs — no extra prompting required by consumers:
- Server instructions (delivered via the MCP initialize handshake): what the mailbox is, the standard workflows (inspect, e2e wait-then-verify, template QA, attachments), search syntax, and warnings — clients inject these into the model's context automatically.
- Tool annotations: the ten read-only tools carry
readOnlyHint,delete_messagescarriesdestructiveHint— clients can auto-approve reads and gate deletes. - MCP prompts:
verify_email(wait for a message matching a query, then verify structure, links and client compatibility against an optional checklist) andinspect_mailbox(summarize mailbox state, flag anomalies). Claude Code exposes these as slash commands.
Claude Code plugin
The easiest way to use this from Claude Code — one install, prompted for your endpoint and token, no manual MCP or skill setup:
/plugin marketplace add CeDJeY/mailpit-mcp-server
/plugin install mailpit@cedjey
On enable, Claude Code asks for:
- Mailpit MCP endpoint URL — e.g.
http://your-host:3000/mcp - Bearer token — the server's
MCP_AUTH_TOKEN(leave empty if auth is disabled)
The plugin bundles the MCP connection plus two skills:
verify-email— the verification workflow: wait for the email, inspect structure, extract links safely, check HTML compatibility, report a PASS/FAIL verdict.route-mail-to-mailpit— the prerequisite the verification depends on: configure the application under test to send SMTP to Mailpit, prove it with a canary email, and flag configurations that could leak test mail to real SMTP providers.
Plugin source lives in plugin/; non-plugin users can copy the skills into their project's .claude/skills/ manually.
Configuration
| Variable | Default | Description |
|---|---|---|
MAILPIT_URL |
http://localhost:8025 |
Base URL of the Mailpit instance |
MCP_TRANSPORT |
http |
http (network service) or stdio (spawned by client) |
MCP_HTTP_PORT |
3000 |
Port for the HTTP endpoint |
MCP_AUTH_TOKEN |
(empty = auth off) | When set, /mcp requires Authorization: Bearer <token> |
MAILPIT_AUTH_USER / MAILPIT_AUTH_PASS |
(none) | Basic-auth credentials if Mailpit's API is protected |
MAILPIT_SMTP_ADVERTISE |
(none) | SMTP host:port applications must send to (as reachable by them); exposed to agents via get_mailbox_info (SMTPEndpoint) and the server instructions so they can verify app mail routing |
The HTTP mode also serves GET /healthz (no auth) for container healthchecks.
Security notes
Mailpit captures every email your app sends — treat the MCP endpoint as sensitive. Set MCP_AUTH_TOKEN whenever the port is reachable beyond localhost, and prefer keeping it off the public internet (private network, VPN, or an authenticated tunnel). delete_messages is destructive; scope access accordingly.
Development
npm ci
MAILPIT_URL=http://localhost:8025 npm start # HTTP mode on :3000
node mcp-test.js # e2e check against a running server
mcp-test.js honours MCP_URL and MCP_AUTH_TOKEN.
License
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.