Ruflo MCP Flaw Lets Unauthenticated Attackers Run Commands and Poison AI Memory
Ruflo MCP Flaw Lets Unauthenticated Attackers Run Commands and Poison AI Memory — [https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhCgiIKxPiETcU1yI
What’s new: A critical vulnerability (CVE-2026-59726, CVSS score: 10.0) has been discovered in Ruflo, an open-source agent meta-harness for AI systems, allowing unauthenticated remote code execution. The flaw affects all versions prior to 3.16.3, exposing 233 tools through an unauthenticated Model Context Protocol (MCP) bridge. Attackers can exploit this to execute commands, steal API keys, and manipulate AI memory.
Who’s affected
All users of Ruflo versions before 3.16.3 are at risk, particularly those with default configurations that expose the MCP bridge on port 3001 to the network.
What to do
- Immediately close firewall ports 3001 and 27017.
- Rotate all LLM API keys.
- Audit the AgentDB pattern store for unauthorized entries.
- Check MongoDB for signs of tampering.
- Upgrade to Ruflo version 3.16.3 or later to mitigate the vulnerability.
