CVE-2026-134553

Wiro MCP upload URLs could reach private network services

Wiro · Wiro-MCP · upload_file

First documented disclosure · Review checkpoint

Summary

A tool call could make the local MCP server fetch a private-network URL and upload the response to Wiro.

Severity and exploitation evidence

Medium · 5.3/10 · CVSS 4.0. Severity source.

Exploitation not established: The reviewed sources document a weakness or controlled demonstration; malicious exploitation against users is not established.

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

The upload URL lacked private-address and redirect validation.

Conditions required

A locally running stdio MCP server has a Wiro API key and can reach the target. The user or agent invokes upload_file with a supplied URL; untrusted content may influence an agent.

Potential impact

Private-network content could be uploaded to the user’s own Wiro account with the local machine’s network reach.

Affected versions and configurations

Versions through 1.3.0 according to the CNA.

Fix and mitigation status

1.3.1; the correction validates public HTTP(S) destinations at socket DNS resolution and at each redirect.

Practical next steps

Upgrade to 1.3.1, limit MCP-server network reach and review uploads made from untrusted URLs.

Operating lesson

Validate each actual network connection and redirect for tools that import remote content.

Evidence and limits

The public 1.3.1 security release is October 16; the researcher report is October 17; CNA ingestion is November 22. VulDB assigns CVSS 4.0 5.3 and 3.1 4.3; the researcher/vendor describe Low severity under the local stdio trust model. The report’s PoC models URL behavior rather than proving a complete attack. Although file:// passed the old syntactic check, Node fetch rejected it, so direct local-file disclosure was not demonstrated. The full correction also bounds redirects, body size and duration. No external attacker-controlled exfiltration account is established.

Sources