Technical impact
The vulnerability lies in the Model Context Protocol Python SDK's handling of HTTP redirects. When follow_redirects is enabled, the SDK follows redirects to a different origin and forwards any custom credential headers supplied by the caller. For 307 or 308 redirection responses, it also forwards the request body. An attacker who can influence the redirect destination of a trusted endpoint can capture the forwarded credentials and request body, exposing sensitive information. The weakness is captured as information disclosure (CWE-200, CWE-201).
Affected Systems
All releases of the Model Context Protocol Python SDK from version 1.8.0 up to, but not including, 1.30.0, and any 2.x releases older than 2.2.0, are susceptible. The SDK components affected include streamable_http_client, streamablehttp_client, sse_client, ClientSessionGroup, OAuth, and the create_mcp_http_client constructor. The vulnerability is fixed starting with SDK version 1.30.0 and 2.2.0.
Risk and Exploitability
The CVSS score of 5.9 classifies the issue as moderate severity. The EPSS score is less than 1%, indicating a low likelihood of exploitation. The vulnerability is not currently listed in the CISA Known Exploited Vulnerabilities catalog. Successful exploitation requires an attacker to control or influence a redirect response from a trusted server endpoint; once achieved, the attacker can obtain the forwarded headers and body. Because the attack path involves manipulating HTTP redirects, the environmental prerequisites are modest but the impact of disclosed credentials makes it a concern for organizations using the SDK without additional safeguards.
Github GHSA