First Malicious MCP Server Found in npm Postmark-MCP Package

Malicious npm Package postmark-mcp Steals Emails via Malicious MCP Server Discovered in Supply Chain Attack

  • Security researchers detected the first known malicious Model Context Protocol (MCP) server in the wild on September 17, 2025.
  • A fake npm package named postmark-mcp imitating a legitimate library was found to steal emails by copying them to an external server.
  • The compromised package was uploaded by developer phanpak, was downloaded over 1,600 times, and later removed from the npm repository.
  • The attack added a BCC line forwarding emails to “phan@giftshop[.]club,” exposing sensitive communications.
  • Developers using this package are advised to remove it, change exposed credentials, and check for unauthorized email forwarding.

On September 17, 2025, Cybersecurity researchers uncovered the first real-world case of a malicious Model Context Protocol (MCP) server embedded in an npm package called postmark-mcp. The package, uploaded by developer phanpak, copied emails sent through the MCP service to a personal server without user consent, creating significant supply chain risks.

- Advertisement -

The MCP server is intended to help users send emails, manage templates, and track campaigns using AI assistants. The legitimate library is available on GitHub and can be accessed through Postmark Labs. However, version 1.0.16 of the npm clone, released on September 17, 2025, included a malicious change that silently forwarded all emails to “phan@giftshop[.]club” by adding a blind carbon copy (BCC).

Phanpak uploaded the fake package on September 15, 2025, and it attracted around 1,643 downloads before being removed from the npm repository. Koi Security CTO Idan Dardikman stated, “Since version 1.0.16, it’s been quietly copying every email to the developer’s personal server.” He emphasized the simplicity of the backdoor and the broad impact it could have by stealing thousands of emails.

MCP servers operate with high trust and permissions inside development toolchains, handling sensitive data like password resets and customer communications. Security company Snyk noted that the backdoor was designed specifically to harvest emails from agentic workflows relying on the MCP server. They highlighted the risks involved due to the elevated privileges and data sensitivity managed by MCP servers.

Developers who installed the compromised package are advised to remove postmark-mcp from their projects immediately. They should rotate any credentials that might have been exposed and review email logs for any unauthorized BCC to the reported domain. This incident illustrates ongoing threats from malicious actors exploiting trust within open-source and emerging ecosystems.

- Advertisement -

✅ Follow BITNEWSBOT on Telegram, Facebook, LinkedIn, X.com, and Google News for instant updates.

Previous Articles:

- Advertisement -

Latest News

Bitcoin Surges as Fed Shift Sparks Market Panic and Gold Rally

Bitcoin prices rebounded after a recent downturn, amid growing speculation about Federal Reserve and...

UGM Pilots Blockchain Credentials, Digital Wallets for 60,000 Students

Universitas Gadjah Mada in Indonesia will launch digital, verifiable credentials for students, starting with...

UK Seizes £5.5B Bitcoin in Largest Crypto Fraud Bust

A Chinese woman pleaded guilty to involvement in a cryptocurrency fraud after £5.5 billion...

Trader Turns $215 Into $1.06M With Pepe (PEPE) Memecoin Surge

PEPE reached an all-time high of $0.00002803 in December 2024.An investor turned a $215...

OpenAI’s ChatGPT Drives $4.3B H1 Revenue Amid Rapid AI Growth

OpenAI reported $4.3 billion in revenue for the first half of 2025, outpacing its...
- Advertisement -

Must Read

8 Best Crypto Debit Cards For Spending Your Digital Tokens

What are | How we chose | Best crypto debit cards | Binance Card? | FAQ | Final WordsCrypto debit cards have transformed how...