Executive brief
n8n is a workflow automation platform that allows users to connect APIs and services through visual workflows. When LLM sub-nodes (like OpenAI or Anthropic) are configured with custom HTTP headers containing credentials, these sensitive values are displayed as masked in the UI but written in plaintext to execution data stored in the database. Any authenticated user viewing execution records can extract exposed API keys and secrets, which can persist indefinitely if the data is exported or archived.
Technical details
The vulnerability is a credential exposure issue (CWE-532) affecting n8n versions before 1.123.64, 2.29.8, and 2.30.1. Custom HTTP headers configured in LLM sub-node credentials (OpenAI, Anthropic, Lemonade) are properly masked in the web UI but are written unencrypted to workflow execution records. An authenticated attacker with access to execution data can retrieve header values containing API keys or secrets; because execution data persists in the database and can be exported, compromise is not limited to runtime exposure. The attack requires authentication and access to the execution logs, but no user interaction beyond navigating to affected workflow history. Patches are available in versions 1.123.64, 2.29.8, and 2.30.1.
Affected products
- n8n n8n < 1.123.64, < 2.29.8, < 2.30.1
Timeline
- 2026-07-08: disclosed: Original advisory GHSA-89gh-3pgc-v5h2 published
- 2026-07-22: disclosed: Duplicate advisory GHSA-fmvg-vhqq-r2mj published and later withdrawn