Executive brief
Langflow is a tool used to build and deploy AI-powered agents and workflows. A security flaw was identified where clicking the logout button failed to actually terminate the user's session, leaving them logged in. This poses a risk in shared computing environments, where a subsequent user could gain full access to the previous user's account, data, and AI workflows.
Technical details
An insufficient session expiration vulnerability (CWE-613) exists in Langflow's logout mechanism. The root cause is twofold: the backend `/logout` endpoint attempted to delete authentication cookies without matching the original security parameters (HttpOnly, SameSite, Secure, and domain), causing browsers to ignore the deletion request; additionally, the frontend failed to explicitly clear these cookies. Consequently, `access_token_lf` and `refresh_token_lf` remained active in the browser. An attacker with physical access to a shared machine can refresh the page after a user "logs out" to resume the session with high confidentiality and integrity impact. This is resolved in version 1.7.0 by ensuring cookie parameter parity during deletion and updating the frontend auth store.
Affected products
- langflow-ai Langflow < 1.7.0
Timeline
- 2025-11-07: other: Fixes submitted via pull requests
- 2025-11-12: patched: Fixes merged into main branch
- 2026-06-19: advisory: GitHub Security Advisory published
- 2026-06-23: disclosed: CVE published to NVD