CVE-2026-15750 Overview
CVE-2026-15750 is a Server-Side Request Forgery (SSRF) vulnerability in the mastergo-design/mastergo-magic-mcp project up to version 0.2.0. The flaw resides in the z.string function within src/tools/get-component-link.ts, part of the mcp__getComponentLink component. Attackers can manipulate the url argument to coerce the server into issuing arbitrary outbound requests. The issue is remotely exploitable and a public exploit is available. The maintainers received an issue report about the problem but had not responded at the time of publication.
Critical Impact
Remote attackers can abuse the mcp__getComponentLink tool to send crafted requests from the MCP server, potentially reaching internal services, cloud metadata endpoints, or other resources normally unreachable from the internet.
Affected Products
- mastergo-design/mastergo-magic-mcp versions up to and including 0.2.0
- Deployments exposing the mcp__getComponentLink MCP tool
- Any integration embedding the vulnerable get-component-link.ts handler
Discovery Timeline
- 2026-07-14 - CVE-2026-15750 published to NVD
- 2026-07-15 - Last updated in NVD database
Technical Details for CVE-2026-15750
Vulnerability Analysis
The vulnerability is a Server-Side Request Forgery classified under [CWE-918]. The mcp__getComponentLink tool exposed by the mastergo-magic-mcp Model Context Protocol (MCP) server accepts a url argument that is validated only as a string via z.string. No allowlist, scheme restriction, or destination validation is applied before the server performs the outbound request. An attacker supplying a crafted URL can direct the server to communicate with arbitrary hosts, including internal-only endpoints reachable from the MCP server's network position.
Root Cause
The root cause is missing validation of the user-supplied url parameter in src/tools/get-component-link.ts. The Zod schema z.string() confirms the value is a string but does not enforce a URL format, restrict schemes, or block private and link-local address ranges. This permits requests to internal IPs, localhost, cloud metadata services, and non-HTTP schemes.
Attack Vector
Exploitation requires network access to the MCP server and user interaction to invoke the tool through an MCP client. An attacker crafts a prompt or tool invocation that supplies a malicious url, such as one pointing to http://169.254.169.254/ for cloud metadata or an internal admin interface. The server issues the request under its own identity and returns the response to the caller. See the GitHub Issue Tracker for the public report describing the exploitation pattern.
Detection Methods for CVE-2026-15750
Indicators of Compromise
- Outbound HTTP requests from the MCP server to RFC1918 addresses, 127.0.0.0/8, or 169.254.169.254
- Invocations of mcp__getComponentLink with url values referencing non-public hostnames or unusual schemes
- Unexpected access log entries on internal services originating from the MCP server host
Detection Strategies
- Instrument the get-component-link.ts handler to log every resolved destination host before the request is issued
- Deploy egress monitoring that flags MCP server traffic to internal or metadata endpoints
- Correlate MCP tool invocation telemetry with network flow data to identify anomalous destinations
Monitoring Recommendations
- Alert on any DNS resolution or TCP connection from the MCP process to private address space
- Monitor for spikes in mcp__getComponentLink invocations from a single client or session
- Retain full request payloads for the tool to support forensic review of supplied URLs
How to Mitigate CVE-2026-15750
Immediate Actions Required
- Restrict network egress from the MCP server host to only the domains required by the mastergo API
- Disable or gate the mcp__getComponentLink tool until a patched release is available
- Run the MCP server in a network segment that cannot reach internal management interfaces or cloud metadata services
Patch Information
At the time of publication, the maintainers of mastergo-design/mastergo-magic-mcp have not released a fixed version and have not responded to the disclosure. Monitor the GitHub Project Repository and the associated VulDB Vulnerability Details for updates.
Workarounds
- Wrap the url parameter with strict validation that enforces https scheme and an allowlist of mastergo domains
- Resolve the hostname before issuing the request and reject any answer in private, loopback, or link-local ranges
- Place the MCP server behind an outbound proxy that enforces destination policy for all HTTP requests
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
