A vulnerability in AWS aws-api-mcp-server before version 1.5.1 allows incorrect normalization of operation names in its custom security policy enforcement. This flaw may enable an attacker to execute AWS API operations across services that the security policy was intended to deny or require additional approval for, by exploiting operation names containing consecutive capital letters. The issue is fixed in version 1.5.1.
CVE-2026-112137 is a Cross-Site Request Forgery (CSRF) vulnerability in Siemens Mendix Studio Pro 11 affecting specific versions. The MCP Server component does not properly validate Content-Type and Origin headers on incoming HTTP requests, which could allow a malicious web page to trick a user's browser into sending unauthorized requests. This vulnerability affects Mendix Studio Pro 11 versions >=11.15.0 <11.16.0 and >=11.12.6 <11.12.7. A fix is available in versions 11.16.0 and 11.12.7.
A vulnerability has been identified in Mendix Studio Pro 11 (All versions >= V11.15.0 < V11.16.0), Mendix Studio Pro 11.12 (All versions >= V11.12.6 < V11.12.7). The affected products MCP Server does not properly restrict its HTTP listener to the loopback interface, leaving it potentially exposed to remote systems. This could allow an attacker with network access to interact with the MCP Server remotely.
The @tsparticles/mcp-server package's HTTP mode uses a single shared StreamableHTTPServerTransport instance for all clients. The first anonymous client to initialize the MCP endpoint monopolizes this shared transport state, causing subsequent clients to be denied new sessions with an HTTP 400 error. This denial-of-service condition persists until the server process is restarted. The issue affects versions prior to 4.3.2 and has been fixed in version 4.3.2.
The @tsparticles/mcp-server HTTP mode in versions prior to 4.3.1 parses incoming JSON request bodies without local error handling. A malformed JSON request causes the JSON.parse() call to throw an unhandled exception, terminating the Node.js process. This results in a denial of service until the process is externally restarted. Version 4.3.1 fixes this by returning an HTTP 400 response for invalid JSON instead of crashing.
The @tsparticles/mcp-server package contains a vulnerability in its HTTP server mode where the POST /mcp endpoint accumulates incoming request body chunks into memory without size limits. This allows an unauthenticated remote attacker to send an arbitrarily large request body, causing the Node.js process to exhaust memory and crash, resulting in a denial of service. The official Docker image runs this HTTP mode by default, making the vulnerability exploitable in typical container deployments. A fix is available in version 4.3.1 that adds a request body size limit.
CVE-2026-123811 is a vulnerability in dynatrace-mcp versions prior to 2.0.0 where the server started with the --http option accepts JSON-RPC requests without authentication or session validation. This allows network-reachable attackers to invoke certain functions such as execute_dql to read sensitive data like logs and security events, or create_dynatrace_notebook to create tenant-visible notebooks. The issue is fixed in version 2.0.0.
A vulnerability in Dynatrace MCP Server versions prior to 2.0.0 allows improper neutralization of Jinja2 template expressions in workflow creation parameters. This can lead to unintended evaluation of template expressions when workflows run, potentially disclosing event or tenant metadata to attacker-controlled Slack channels. The issue requires either an MCP prompt-injection vector or operator approval of a prompt displaying the template fragment. The vulnerability is fixed in version 2.0.0.
Incorrect Privilege Assignment vulnerability in AtlasAiDev MCP Server for WordPress ai-workflow-automation-ai-agent-hub allows Privilege Escalation.This issue affects MCP Server for WordPress: from n/a through 2.0.1.
fast-agent versions prior to 0.7.21 have an OS command injection vulnerability in the serve component when started with the --shell option. The server listens on all interfaces without authentication on port 8000, allowing unauthenticated remote clients to execute operating system commands as the server user via the /mcp endpoint. Without the --shell option, clients can still access the operator's model and configured MCP tools. This issue is fixed in version 0.7.21.
The Model Context Protocol TypeScript SDK provides TypeScript client and server implementations of the Model Context Protocol. From 2.0.0 until 2.3.0 in @modelcontextprotocol/server, applications that shared one stateless HTTP transport or one McpServer or Server instance among multiple clients could deliver responses and notifications to the wrong client. The affected surface is an MCP server reached over HTTP by multiple clients. The trigger is reuse of one stateless transport or server instance. Messages were associated with a connection using only the JSON-RPC request identifier. SDK clients start request identifiers at zero, so collisions occur in normal operation. Cross-client delivery exposes one client's responses and notifications to another client. The issue requires an HTTP endpoint reachable by more than one client and an application that constructs the server or transport once at startup instead of once per request or session. stdio servers and factory-created per-call servers are not affected. Upgrading causes reused instances to fail and therefore requires an application code change. This issue is fixed in @modelcontextprotocol/server 2.3.0.
CVE-2026-111771 is an incorrect authorization vulnerability in the @modelcontextprotocol SDK affecting server-side bearer authentication. The flaw allows acceptance of a valid access token issued for a different service because the SDK did not enforce token audience validation. An attacker with a token for another service could authenticate to the MCP server and gain unauthorized confidentiality and integrity capabilities. The issue affects multiple packages and versions of the SDK and is fixed in versions 1.32.0, 2.3.0, 2.0.2, and 2.3.1. Upgrading alone is insufficient; operators must enable audience validation and configure the expected token audience to fully mitigate the vulnerability.
CVE-2026-77604 is a path traversal vulnerability in vitali87's code-graph-rag prior to version 0.0.345. The flaw exists in the MCPToolsRegistry.read_file function, which improperly handles attacker-controlled file paths when pagination parameters are used, allowing bypass of path validation. This enables unauthenticated remote attackers to read arbitrary UTF-8-readable files accessible to the server process, including sensitive files such as credentials and private keys. The issue is fixed in version 0.0.345.
Rmcp versions prior to 3.9.0 contain a vulnerability in their built-in Streamable HTTP clients where response bodies are buffered in memory without size limits. This affects the reqwest and Unix socket clients, causing them to fully buffer certain HTTP responses, including application/json POST responses and non-2xx error responses. A malicious or compromised MCP server can exploit this by sending an unending response body, leading to memory exhaustion and process termination. The issue is fixed in version 3.9.0 by imposing strict size limits on response bodies and rejecting oversized content.
The 'teamviewer-mcp-server' package version 0.0.1 on PyPI has been identified as malicious by the OpenSSF Package Analysis project. The package is flagged due to its communication with a domain linked to malicious activity. No CVSS score or detailed technical impact is provided.