Connect your AI agents to Postmark

Postmark's MCP Server lets AI tools like Claude Desktop, Cursor, and Windsurf securely access and manage your Postmark account through natural language commands. It connects 24 tools across eight categories — sending, templates, message search, diagnostics, bounces, suppressions, webhooks, and server info — so your AI assistant can handle your email infrastructure conversationally, with no coding required.

• Send emails. Send plain text or HTML to a single address or up to 50 recipients at once. Add CC, BCC, and reply-to. Send up to 500 messages in a single batch call.
• Manage templates. List, create, edit, delete, and validate templates — including layout binding. Prompt your AI to create a template, test it, and send it in the same conversation.
• Diagnose delivery. Ask "Did my email reach this address?" The diagnoseDelivery tool checks message history, suppression status, and bounce history in parallel and gives you a plain-English summary.
• Track stats. Get open rates, click rates, bounce rates, and spam rates. Filter by tag or date range, or drill into platform and client breakdowns.
• Manage bounces and suppressions. Search bounces, retrieve bounce content, reactivate addresses, and add or remove suppression entries — up to 50 at a time.
• Manage webhooks. List, create, and delete webhooks. Webhook URLs must be HTTPS. Optionally set a URL allowlist to restrict which endpoints the server will accept.

The quickest path requires no local clone. Add this code snippet to your MCP client config (Cursor, Claude Desktop, or Windsurf):

{ "mcpServers": { "postmark": { "command": "npx", "args": ["-y", "@activecampaign/postmark-mcp"], "env": { "POSTMARK_SERVER_TOKEN": "your-postmark-server-token", "DEFAULT_SENDER_EMAIL": "your-sender@example.com", "DEFAULT_MESSAGE_STREAM": "outbound" } } } }

Then add your Postmark server token (found in your account under Servers → API Tokens) and start prompting.

Prefer a local install? Clone the repo, run npm install, copy .env.example to .env, and run npm start. Full instructions are in the README on GitHub.

The MCP server works with any tool that supports the Model Context Protocol — including Claude Desktop, Cursor, Windsurf, and a growing ecosystem of AI development tools. The protocol is open-source and designed for extensibility. You can install the server with a single npx command, with no local clone required.

The MCP server is completely free and open-source (MIT licensed). You only need an active Postmark account for sending emails, which follows standard Postmark pricing.

The server is open source (MIT licensed) and designed for extensibility. You can fork the repo, add custom tools, or integrate it into your own AI workflows. A 52-test suite covers unit, offline MCP validation, and end-to-end scenarios — so you can contribute with confidence. Check the GitHub repository for contribution guidelines.

Your Postmark API token is stored locally and never transmitted to any AI provider — the MCP server communicates directly with Postmark's API over HTTPS. Email addresses in logs are masked by default (e.g., a***e@example.com) to reduce PII exposure. Webhook creation requires HTTPS URLs, and you can set a URL allowlist to restrict which endpoints the server will accept. All 24 tools carry read-only, destructive, and idempotency annotations, so MCP clients can surface risk indicators before taking action.