Disclosed weakness · no CVE recorded

A slow Atlassian site could stall every caller of a shared MCP server

mcp-atlassian maintainers · mcp-atlassian · Jira and Confluence MCP server

First documented disclosure · Review checkpoint

Summary

One slow upstream request or a run of failures from one site could interrupt unrelated clients.

Severity and exploitation evidence

Medium · numeric CVSS score not recorded. Severity source.

Exploitation not established: The maintainer reports an internal audit finding. The advisory identifies no malicious use.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

Tool handlers performed synchronous upstream work on the shared event loop and shared circuit-breaker state across sites.

Conditions required

Multiple clients share an HTTP deployment that calls Jira or Confluence.

Potential impact

Other users could lose availability, including users of a healthy upstream site.

Affected versions and configurations

Before 0.24.5.

Fix and mitigation status

0.24.5.

Practical next steps

Upgrade to 0.24.6 or later. Allowlist upstream domains while upgrading.

Operating lesson

Isolate upstream work and failure handling by site on shared tool servers.

Evidence and limits

The maintainer traced this issue to an internal audit and published the patch and advisory November 16. The advisory provides a severity label without a numeric score.

Sources