NewApp Monitoring is liveSign up free
AgentbeamBeta
AI agent vulnerability · Flowise

CVE-2026-40933: Flowise: Authenticated RCE Via MCP Adapters

criticalCVSS 3.1 9.9GHSA-c9gw-hvqq-f33r

Flowise and flowise-components up to 3.0.13 let any authenticated user add a Custom MCP stdio server and run arbitrary commands on the host. The command allowlist checks only the program name, so `npx` with the `-c` argument executes shell code. Upgrade both packages to 3.1.0. The CVSS score is 9.9.

Published Apr 16, 2026 · Analysis Oct 30, 2026

What an attacker can do

The Custom MCP configuration in the Flowise canvas lets a user define an MCP server that runs over stdio, with a command and arguments. Flowise validates this with validateCommandInjection, validateArgsForLocalFileAccess and a list of allowed commands: node, npx, python, python3 and docker. The list checks the command name only. An allowed program can still take arguments that execute code.

The proof of concept sets the command to npx and the arguments to -c and touch /tmp/pwn. npx runs that string in a shell, so any command runs as the Flowise process user. The vulnerable code is in packages/components/nodes/tools/MCP/core.ts.

The attacker needs a Flowise login. Any user can reach the canvas and add a Custom MCP, so a low-privileged or self-registered account is enough. The vector shows a changed scope. From the shell, an attacker can read the API keys, credentials and flow data that Flowise stores, and move into the network around the server. The EPSS score is among the higher ones for this set.

Who is affected

The flowise and flowise-components npm packages, version 3.0.13 and earlier. Check what you run:

npm ls flowise flowise-components

For Docker, check the image tag. Instances where several people hold accounts, or where signup is open, carry the most risk.

flowise

Ecosystem
npm
Affected
<= 3.0.13
Fixed in
3.1.0
Check
npm ls flowise

flowise-components

Ecosystem
npm
Affected
<= 3.0.13
Fixed in
3.1.0
Check
npm ls flowise-components

How to fix it

Upgrade flowise and flowise-components to 3.1.0:

npm install -g [email protected]

The advisory only names the patched version. It does not describe the code change and lists no workaround. Until you upgrade, restrict who can log in to Flowise and who can create or edit Custom MCP stdio configurations. Review existing Custom MCP entries for commands like npx -c. Run Flowise as an unprivileged user in a container, with no cloud credentials in its environment.

How to tell if you were hit

Open the Custom MCP configurations and look for stdio entries where the command is npx, node, python, python3 or docker with arguments such as -c, -e or --eval, or arguments that contain shell commands. On the host, look for child processes of the Flowise Node process, such as sh, curl, wget or bash, and for unexpected files like /tmp/pwn from the proof of concept. Review Flowise logs for MCP server launches you did not configure, and outbound connections from the Flowise host. Beam records the shell commands an agent runs, but a command launched by the Flowise server itself is outside that.

Original advisory

From the GitHub Advisory Database (CC BY 4.0).

Summary

Due to unsafe serialization of stdio commands in the MCP adapter, an authenticated attacker can add an MCP stdio server with an arbitrary command, achieving command execution.

Details

The vulnerability lies in a bug in the input sanitization from the “Custom MCP” configuration in http://localhost:3000/canvas - where any user can add a new MCP, when doing so - adding a new MCP using stdio, the user can add any command, even though your code have input sanitization checks such as validateCommandInjection and validateArgsForLocalFileAccess, and a list of predefined specific safe commands - these commands, for example "npx" can be combined with code execution arguments ("-c touch /tmp/pwn") that enable direct code execution on the underlying OS.

https://github.com/FlowiseAI/Flowise/blob/d848baeb6bd9737a1e7fc912349c45fbdcc7bb38/packages/components/nodes/tools/MCP/core.ts#L223

https://github.com/FlowiseAI/Flowise/blob/d848baeb6bd9737a1e7fc912349c45fbdcc7bb38/packages/components/nodes/tools/MCP/core.ts#L177

https://github.com/FlowiseAI/Flowise/blob/d848baeb6bd9737a1e7fc912349c45fbdcc7bb38/packages/components/nodes/tools/MCP/core.ts#L269

PoC

Create a new Custom MCP and add an "npx -c" command.

{
    "command": "npx",
    "args": [
        "-c",
        "touch /tmp/pwn"
    ]
}

Impact

This is an authenticated arbitrary command execution due to unsanitized input, even though the input is sanitized, more protections should be added in order to close ways for attackers to execute arbitrary commands.

Credited to MosesOX, 13ph03nix.

Questions

Does CVE-2026-40933 need an account?

Yes. The attacker must be authenticated, but any user who can reach the canvas can add a Custom MCP. The CVSS vector requires only low privileges.

Why does the command allowlist not stop it?

It checks only the program name. An allowed program such as npx accepts arguments like -c that run shell code, so the allowlist and the injection checks pass.

Which version fixes it?

3.1.0 for both flowise and flowise-components. Version 3.0.13 and earlier are affected.

References

How this page is made: the data panel comes from the GitHub Advisory Database, CISA’s KEV catalog, the NVD and FIRST EPSS, and updates twice a day. The analysis is written by an AI model from the sources listed above and checked by automated rules before it is published. Check versions against the vendor's advisory before you act. Methodology

More Flowise vulnerabilities

All Flowise entries