A significant security vulnerability has been identified within the Model Context Protocol (MCP) and its associated SDKs, affecting numerous AI clients and developer tools. Initially discovered in the mcp-remote proxy, which facilitates communication between AI clients like Claude Desktop and Cursor and remote MCP servers, the flaw allowed for full remote code execution. While mcp-remote was patched, a subsequent investigation by OX Security revealed similar vulnerabilities embedded within the official MCP SDKs themselves, impacting over 150 million downloads and an estimated 200,000 instances. Anthropic, a key player in the MCP ecosystem, has opted not to alter the protocol's architecture, deeming the behavior expected, which leaves the ecosystem to adapt to this inherent design. AI
IMPACT Exposes millions of developer instances to command injection attacks, necessitating adaptation around a core AI protocol's design.
RANK_REASON Security vulnerabilities in an AI protocol and its SDKs affecting multiple developer tools.
- Anthropic
- Claude Desktop
- Cursor
- Java
- JFrog Ltd
- mcp-remote
- Model Context Protocol
- OX Security
- Python
- Rust
- TypeScript
- Visual Studio Code
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
