UTC
Critical

MCP Path Traversal Allows Full-Filesystem Read and JWT Secret Exfiltration Leading to Account Takeover

CVE-2026-7646 · GHSA-95rw-c7w3-xh7f · published 2026-11-17

Severity
Critical
CVSS score
9.6
Chance of exploitation (EPSS)
Not scored
Fixed in
1.9.1
  • Exploitation No known exploitation
  • Automatable No
  • Technical impact Partial
  • CISA SSVC
What it is

This is an arbitrary server-side file read vulnerability that escalates to full account takeover: 1. Any file readable by the Langflow process can be retrieved — including secret_key, private_key.pem, application configs, and database files. 2. JWT secret exfiltration leads to complete account takeover of all users — forged tokens can be created for any user UUID (including superusers), are non-expiring, and are fully accepted by the server. A single low-privilege account compromises every user on the instance. …

What to do

Upgrade langflow to 1.9.1 or later, and langflow-base to 0.9.1 or later. If you start it from an MCP config with npx or uvx, pin that version so an old cached copy is not used.

This advisory tracks CVE-2026-7646: URL-encoded filename traversal in MCP resources/read, including the possibility of reading the JWT signing secret. The reported path was fixed by PR #12818, merged on April 21, 2026 and included in Langflow 1.9.1, published April 24, 2026 (UTC). The corresponding backend package is langflow-base 0.9.1. The handler rejects decoded traversal sequences and path separators before reading storage; local storage also validates the resolved path. The change adds resource ownership checks. Upgrade recommendation: Follow IBM's published bulletin and upgrade to Langflow 1.11.0 or newer.

Affected
PackageRegistryAffectedFixed in
langflowPyPI< 1.9.11.9.1
langflow-basePyPI< 0.9.10.9.1

The impact, patch and workaround text comes from the advisory, shortened (published by Langflow). Proof-of-concept and reproduction sections and code are left out; the original advisory has them.

Sources: GitHub Advisory Database, reviewed and malware advisories (CC BY 4.0, https://github.com/advisories); advisories published by Cursor (https://github.com/cursor/cursor/security/advisories); advisories published by OpenClaw (https://github.com/openclaw/openclaw/security/advisories); advisories published by vLLM (https://github.com/vllm-project/vllm/security/advisories); advisories published by Open WebUI (https://github.com/open-webui/open-webui/security/advisories); advisories published by n8n (https://github.com/n8n-io/n8n/security/advisories); advisories published by Flowise (https://github.com/FlowiseAI/Flowise/security/advisories); advisories published by Langflow (https://github.com/langflow-ai/langflow/security/advisories); advisories published by the MCP Python SDK project (https://github.com/modelcontextprotocol/python-sdk/security/advisories); advisories published by MLflow (https://github.com/mlflow/mlflow/security/advisories); CISA Known Exploited Vulnerabilities catalog (CC0). Data as of 2026-11-22. ShieldMap describes how each problem works and how to fix it; it does not publish exploit code. Report a correction to [email protected].