CVE-2026-130796

Flowise public chatflow responses could reveal MCP access tokens

FlowiseAI · Flowise

First documented disclosure · Review checkpoint

Summary

A public chatflow response could disclose the bearer token protecting its MCP endpoint.

Severity and exploitation evidence

High · 8.8/10 · CVSS 4.0. Severity source.

Exploitation not established: Malicious exploitation remains unverified in the reviewed sources.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

The unauthenticated chatflow handler returned secret MCP configuration.

Conditions required

The attacker knows a chatflow identifier; that flow is marked isPublic=true and its MCP server is enabled.

Potential impact

The token can authorize chatflow calls that bypass API-key protection and consume the owner’s model quota.

Affected versions and configurations

3.1.3 through 3.1.4, according to the CNA.

Fix and mitigation status

A corrected release remains unverified.

Practical next steps

Restrict access to affected public chatflow and MCP routes, disable exposed flows, and rotate disclosed tokens before reopening access.

Operating lesson

Exclude authentication material from every public representation of an agent workflow.

Evidence and limits

The researcher file first appeared in public commit 375b58f on November 17 at 14:58:38 UTC, before the prior checkpoint. Its June 9 report was explicitly private. The researcher describes versions through 3.1.3; the current CNA specifies 3.1.3 through 3.1.4. Public demonstrations do not establish malicious use or a shipped correction.

Sources