MoLing MCP server lets web pages control it without login
If you run MoLing's network mode and open a malicious website, that site can quietly send commands to MoLing on your computer, which may read or change your files or run programs. Anyone on the internet who lures you to the page can do this
| Severity | CriticalCVSS 9.6 of 10 |
|---|---|
| Being exploited? | UnclearNo reports either way yet |
| Fix available? | YesVersion 0.5.0. · since 30 Oct 2026 |
| How sure is this? | Official sourceBacked by an official source |
How the attack works simplified
- 1Attacker builds pagePage hosts cross-origin EventSource script
- 2Victim visits pageBrowser opens connection to local /sse
- 3sessionId leakedEndpoint event read due to wildcard CORS
- 4Forged tool callsSent to /message with stolen sessionId
- 5Files and commandsDepends on tools configured
- ResultAttacker runs actions on victim machine
What happened
MoLing versions before 0.5.0 running in SSE mode (-l) expose /sse and /message HTTP endpoints without authentication and send Access-Control-Allow-Origin: * via the mcp-go dependency. A malicious page can read the sessionId from the /sse endpoint event and then send unauthorized MCP tool calls to /message, even when bound to loopback.
Are you affected?
You are affected if you run MoLing before version 0.5.0 in SSE mode using the -l flag, regardless of whether it listens only on localhost.
MoLing versions prior to 0.5.0 running in SSE mode (-l).
What to do
- Update MoLing to version 0.5.0 or later
- Stop running MoLing in SSE mode (-l) until updated
- Avoid browsing untrusted websites while MoLing SSE mode is running
Timeline
- 30 Oct 2026, 21:17 UTCPublicly disclosed
- 30 Oct 2026, 21:17 UTCReported by NVD · NVD
- 30 Oct 2026, 21:20 UTCPicked up by Zero Radar
Sources 1 report, oldest first
Official vendor, government CERT or advisory database
- CVE-2026-62863NVD · 30 Oct 2026, 21:17 UTC
Technical details
- ID
- CVE-2026-62863
- CVSS
- 9.6 / 10
- Disclosed
- 30 Oct 2026, 21:17 UTC
- On Zero Radar
- 30 Oct 2026, 21:20 UTC (3 min after disclosure)
- Last change
- 30 Oct 2026, 21:20 UTC
The plain-language parts and the attack diagram were written by an AI model (anthropic/claude-haiku-5.5) from the sources above. Facts like versions, exploitation and dates come from the sources themselves. How Zero Radar works