Network‑AI is a TypeScript/Node.js orchestrator that in versions 5.7.1 and earlier exposed its SSE MCP server to unauthenticated access because the default secret was empty. The isAuthorized() routine returned true when the secret was blank, allowing any client to invoke all 22 MCP tools, including privileged operations such as agent_spawn and token_* management. This flaw means an attacker can run arbitrary orchestration commands on the host without credentials, effectively gaining remote code‑execution capability.

The vulnerability affects Jovancoding’s Network‑AI product, specifically versions 5.7.1 and earlier. The issue was addressed in release 5.7.2. No other vendor or product is listed, so only users running the affected Network‑AI versions are at risk.

The CVSS score of 9.1 classifies the flaw as critical. EPSS <1% indicates a low expected exploitation probability at the time of assessment, and the vulnerability is not listed in the CISA KEV catalog. Nevertheless, because the flaw permits unauthenticated remote command execution, an attacker with network reach to the SSE endpoint can immediately gain full control. The attack requires only a simple HTTP request to the server’s SSE socket, so any external process such as curl, SSRF, or a remote control system can abuse it if the server is reachable.