The vulnerability arises from MCP Atlassian’s use of the module-level requests.get function instead of the internally hooked session during Jira permission lookup, allowing a caller-provided Jira URL to be redirected to an internal address. This bypasses the redirect checks that were added for a prior CVE, exposing an internal network endpoint to an external attacker. The flaw therefore enables a server‑side request forgery that can leak internal services or data, compromising confidentiality, integrity, and potentially availability of internal resources.

Sooperset’s MCP Atlassian, a Model Context Protocol server that interfaces with Atlassian Confluence and Jira, is vulnerable in all releases earlier than 0.22.0. The fix was deployed in version 0.22.0, making that and later releases safe.

The CVSS score of 5.3 indicates a moderate severity. EPSS information is not available and the vulnerability is not listed in the CISA KEV catalog, suggesting a lower exploitation probability overall. The likely attack vector is a remote attacker manipulating a publicly reachable Jira URL to redirect the unhooked request to a private internal address. If successful, the attacker can access any internal service reachable from the MCP Atlassian server, potentially exposing sensitive data or facilitating further lateral movement. Because the attack requires only supply of an unsafe URL, the risk to environments lacking strict URL validation or network segmentation is elevated, warranting attention.