A recently reported security issue involves a path traversal vulnerability that can lead to a supply chain compromise affecting MCP server hosting environments. This vulnerability allows attackers to access unauthorized files and potentially manipulate server hosting infrastructure, escalating to a broader supply chain attack. Although no specific affected versions or patches are detailed, the threat is considered medium severity due to its potential impact and exploitation complexity. There are no known exploits in the wild yet, and technical details remain limited, primarily sourced from a Reddit NetSec discussion linking to a GitGuardian blog post. European organizations using MCP server hosting or related supply chain components should be vigilant. Mitigation requires thorough code audits for path traversal flaws, strict input validation, and enhanced supply chain security practices. Countries with significant cloud infrastructure and supply chain dependencies, such as Germany, France, and the UK, are more likely to be impacted. The threat's medium severity reflects moderate impact potential with some exploitation difficulty and limited current exposure. Defenders should prioritize monitoring for unusual file access patterns and supply chain anomalies while awaiting further technical disclosures and patches.
Dify is an LLM application development platform. In Dify versions through 1.9.1, the MCP OAuth component is vulnerable to cross-site scripting when a victim connects to an attacker-controlled remote MCP server. The vulnerability exists in the OAuth flow implementation where the authorization_url provided by a remote MCP server is directly passed to window.open without validation or sanitization. An attacker can craft a malicious MCP server that returns a JavaScript URI (such as javascript:alert(1)) in the authorization_url field, which is then executed when the victim attempts to connect to the MCP server. This allows the attacker to execute arbitrary JavaScript in the context of the Dify application.
A credential weakness has been identified in the Model Context Protocol (MCP) server, raising security concerns. The weakness relates to how credentials are managed or validated within the MCP, potentially allowing unauthorized access if exploited. While no known exploits are currently in the wild, the issue is rated as medium severity due to potential risks to confidentiality and integrity. The threat was recently reported by ReversingLabs and discussed minimally on Reddit's InfoSecNews community. European organizations using MCP or related systems should be aware of this vulnerability and assess their exposure. Mitigation involves reviewing credential management practices, enforcing strong authentication, and monitoring for unusual access patterns. Countries with significant industrial, governmental, or technological deployments of MCP-related infrastructure are more likely to be affected. Given the lack of detailed technical data and no public exploits, the threat is assessed as medium severity. Defenders should prioritize credential security audits and prepare for potential future exploit attempts.
A vulnerability was determined in samanhappy MCPHub up to 0.9.10. This affects an unknown part of the file src/controllers/serverController.ts of the component MCPRouter Service. This manipulation of the argument baseUrl causes server-side request forgery. The attack may be initiated remotely. The exploit has been publicly disclosed and may be utilized. The vendor was contacted early about this disclosure but did not respond in any way.
vet is an open source software supply chain security tool. Versions 1.12.4 and below are vulnerable to a DNS rebinding attack due to lack of HTTP Host and Origin header validation. Data from the vet scan sqlite3 database may be exposed to remote attackers when vet is used as an MCP server in SSE mode with default ports through the sqlite3 query MCP tool. This issue is fixed in version 1.12.5.
ADB MCP Server is a MCP (Model Context Protocol) server for interacting with Android devices through ADB. In versions 0.1.0 and prior, the MCP Server is written in a way that is vulnerable to command injection vulnerability attacks as part of some of its MCP Server tool definition and implementation. This issue has been patched via commit 041729c.
A vulnerability was detected in sequa-ai sequa-mcp up to 1.0.13. This affects the function redirectToAuthorization of the file src/helpers/node-oauth-client-provider.ts of the component OAuth Server Discovery. Performing manipulation results in os command injection. Remote exploitation of the attack is possible. The exploit is now public and may be used. Upgrading to version 1.0.14 is able to mitigate this issue. The patch is named e569815854166db5f71c2e722408f8957fb9e804. It is recommended to upgrade the affected component. The vendor explains: "We only promote that mcp server with our own URLs that have a valid response, but yes if someone would use it with a non sequa url, this is a valid attack vector. We have released a new version (1.0.14) that fixes this and validates that only URLs can be opened."
The mcp-database-server (MCP Server) 1.1.0 and earlier, as distributed via the npm package @executeautomation/database-server, fails to implement adequate security controls to properly enforce a "read-only" mode. This vulnerability affects only the npm distribution; other distributions are not impacted. As a result, the server is susceptible to abuse and attacks on affected database systems such as PostgreSQL, and potentially others that expose elevated functionalities. These attacks may lead to denial of service and other unexpected behaviors.
hackmd-mcp is a Model Context Protocol server for integrating HackMD's note-taking platform with AI assistants. From 1.4.0 to before 1.5.0, hackmd-mcp contains a server-side request forgery (SSRF) vulnerability when the server is run in HTTP transport mode. Arbitrary hackmdApiUrl values supplied via the Hackmd-Api-Url HTTP header or a base64-encoded JSON query parameter are accepted without validation, allowing attackers to redirect outbound API requests to internal network services, access internal endpoints, perform network reconnaissance, and bypass network access controls. The stdio transport mode is not affected because it only accepts stdio requests. The issue is fixed in version 1.5.0, which enforces allowed endpoints and supports the ALLOWED_HACKMD_API_URLS environment variable. Users should update to 1.5.0 or later or apply documented mitigations such as switching to stdio mode, restricting outbound network access, or filtering the Hackmd-Api-Url header and related query parameter via a reverse proxy.
feiskyer mcp-kubernetes-server through 0.1.11 allows OS command injection, even in read-only mode, via /mcp/kubectl because shell=True is used. NOTE: this is unrelated to mcp-server-kubernetes and CVE-2025-53355.
feiskyer mcp-kubernetes-server through 0.1.11 does not consider chained commands in the implementation of --disable-write and --disable-delete, e.g., it allows a "kubectl version; kubectl delete pod" command because the first word (i.e., "version") is not a write or delete operation.
CVE-2025-58444 is a high-severity cross-site scripting (XSS) vulnerability in the modelcontextprotocol (MCP) Inspector local development tool. The flaw occurs when the Inspector renders a redirect URL from a remote MCP server without proper neutralization of encoded URI schemes. Connecting to an untrusted server can allow crafted redirects to inject scripts into the Inspector context, potentially leading to arbitrary command execution on the developer's machine. Version 0.16.6 of the Inspector includes hardened URL handling and validation that prevents this script execution.
@akoskm/create-mcp-server-stdio is an MCP server starter kit that uses the StdioServerTransport. Prior to version 0.0.13, the MCP Server is written in a way that is vulnerable to command injection vulnerability attacks as part of some of its MCP Server tool definition and implementation. The MCP Server exposes the tool `which-app-on-port` which relies on Node.js child process API `exec` which is an unsafe and vulnerable API if concatenated with untrusted user input. Version 0.0.13 contains a fix for the issue.
Markdownify is a Model Context Protocol server for converting almost anything to Markdown. Versions below 0.0.2 contain a command injection vulnerability, caused by the unsanitized use of input parameters within a call to child_process.exec, enabling an attacker to inject arbitrary system commands. Successful exploitation can lead to remote code execution under the server process's privileges. The server constructs and executes shell commands using unvalidated user input directly within command-line strings. This introduces the possibility of shell metacharacter injection (|, >, &&, etc.). This issue is fixed in version 0.0.2.
The npm package wealthsimple-mcp-server version 99.0.0 contains malicious code that installs a dependency from a custom domain. This dependency executes an information stealer during installation, targeting sensitive tokens related to npm, git, and CI/CD environments.
A security flaw has been discovered in AiondaDotCom mcp-ssh up to 1.0.3. Affected by this issue is some unknown functionality of the file server-simple.mjs. Performing manipulation results in command injection. The attack can be initiated remotely. Upgrading to version 1.0.4 and 1.1.0 can resolve this issue. The patch is named cd2566a948b696501abfa6c6b03462cac5fb43d8. It is advisable to upgrade the affected component.
LSTM-Kirigaya's openmcp-client is a vscode plugin for mcp developer. Prior to version 0.1.12, when users on a Windows platform connect to an attacker controlled MCP server, attackers could provision a malicious authorization server endpoint to silently achieve an OS command injection attack in the open() invocation, leading to client system compromise. This issue has been patched in version 0.1.12.
Cherry Studio is a desktop client that supports for multiple LLM providers. In version 1.5.1, a remote code execution (RCE) vulnerability exists in the Cherry Studio platform when connecting to streamableHttp MCP servers. The issue arises from the server’s implicit trust in the oauth auth redirection endpoints and failure to properly sanitize the URL. This issue has been patched in version 1.5.2.