The vulnerability causes full arguments and JSON‑RPC parameters of authenticated MCP tool calls to be written to the server logs before any redaction is performed. When those arguments contain credential material—such as bearer tokens, OAuth secrets, per‑tenant API keys, webhook authentication headers, or arbitrary secret‑bearing payloads—the raw values are persisted in logs that may be accessible outside the trusted request boundary. This exposure can lead to disclosure of sensitive authentication information and other secrets.

The affected product is n8n‑mcp, a computational‑modeling platform for AI assistant integration built by the vendor identified as “czlonkowski:n8n‑mcp.” All releases prior to version 2.47.13 are impacted; version 2.47.13 and later contain the patch that eliminates the logging of full tool‑call arguments in HTTP transport mode.

The CVSS score of this issue is 4.3, indicating a moderate severity. Because the flaw requires that the caller be authenticated and the application is running in HTTP mode, the attack vector is limited to legitimate users with valid credentials. However, in environments where logs are exported to SIEMs, shared log storage, or other external systems, the exposed data could be collected and examined by adversaries or third‑party operators, resulting in potential compromise of token or API key security. The EPSS score is unavailable and the vulnerability is not listed in CISA’s KEV catalog, suggesting that exploitation is not widely observed yet, but the potential for data leakage remains significant if logs are not properly protected.