PT-2026-33361

CVE-2026-40933

·

Published

2026-04-16

·

Updated

2026-09-18

Description by

CVSS 3.1
9.9
Critical
MetricsAV:N/AC:L/PR:L/UI:N/S:C/C:H/I:H/A:H
Name of the Vulnerable Software and Affected Versions Flowise versions prior to 3.1.0
Description An issue exists in the MCP adapter due to unsafe serialization of stdio commands, allowing an authenticated attacker to achieve command execution on the underlying operating system. The flaw is located in the input sanitization of the Custom MCP configuration at the endpoint "http://localhost:3000/canvas". Although the system employs sanitization checks such as validateCommandInjection() and validateArgsForLocalFileAccess(), certain predefined safe commands like "npx" can be combined with malicious arguments (e.g., -c) to bypass these protections. Attackers can exploit this by creating a Custom MCP Tool with a malicious command configuration or by manipulating users into importing a specially crafted chatflow. This allows for the execution of arbitrary code, potentially granting access to sensitive data and connected services. The Model Context Protocol (MCP) is a standard that allows AI Large Language Models to interface with external agents.
Recommendations Update to version 3.1.0 or later.
Exploitability
Exploit
Fix
Impacts
Vendors
Affected products
Related identifiers
GHSA-C9GW-HVQQ-F33R
Found an issue in the description? Have something to add? Feel free to write us 👾