CVE-2026-81098 is a critical vulnerability in the Telnyx MCP server where the HTTP transport was exposed on all interfaces without requiring authentication. This allowed unauthenticated callers to access critical functions and dispatch requests using the server's stored credentials, including API keys and secrets. The vulnerability affects versions up to and including 6.83.0. The current code has been updated to restrict the host to loopback and enforce API key authentication via middleware.
CVE-2026-81097 is a high-severity OS command injection vulnerability in the maquina-app rails-mcp-server affecting versions 1.4.0 through 1.6.0. The vulnerability arises because the execute_ruby tool, intended as a read-only Ruby sandbox, uses a denylist pattern and method replacements to restrict process spawning. However, the pseudo-terminal library's spawn entry points are not covered by this denylist or replacements, allowing an attacker to execute arbitrary commands as the server's user. Version 1.6.1 introduces restrictions on permitted requires and blocks dynamic dispatch to execution entry points, and version 2.0.0 removes the vulnerable tool entirely.
CVE-2026-81095 is a high-severity vulnerability in timescale's pg-aiguide prior to version 0.5.1. The issue arises because the MCP HTTP transport was started without enabling the host allow-list protection provided by the underlying SDK. This omission allows an attacker to craft a request with a controlled host name that the server accepts, enabling potential DNS rebinding attacks through a victim's browser. The vulnerability is fixed in version 0.5.1 by explicitly enabling the DNS-rebinding-protection option.
CVE-2026-81093 is a Server-Side Request Forgery (SSRF) vulnerability in the apify actors-mcp-server prior to version 0.9.12. The vulnerability arises because the get-html-skeleton tool fetches a URL supplied by the caller after only validating its syntax, without verifying the hostname or IP address. This allows attackers to make the server request internal endpoints, including cloud instance metadata services, potentially exposing sensitive information such as instance credentials. The vulnerable tool was removed in version 0.9.12.
CVE-2026-81486 is a path traversal vulnerability in the bsmi021 mcp-file-context-server version 1.0.0. The flaw exists in the read_context function of the src/index.ts file within the Path Resolution component. An attacker can remotely manipulate the path argument to perform path traversal. Exploit code is publicly available, but the vendor has not yet responded or issued a fix.
CVE-2026-81421 is a server-side request forgery (SSRF) vulnerability in ddfourtwo sentry-selfhosted-mcp version 0.4.0. The flaw exists in an unknown function of the raw_sentry_api component where manipulation of the 'endpoint' argument allows remote attackers to perform SSRF. The vulnerability has a CVSS 4.0 base score of 6.9, indicating medium severity. Exploit code has been publicly released, but there is no vendor response or patch available at this time.
CVE-2026-80347 is a Server-Side Request Forgery (SSRF) vulnerability in the kazuph mcp-fetch tool. The vulnerability arises because the SSRF guard does not correctly handle IPv6 literals enclosed in brackets, causing the private address checks to be bypassed. This allows an attacker who can influence fetch targets to make requests to internal or private network addresses, potentially exposing sensitive internal responses to the attacker.
CVE-2026-53965 is a medium severity vulnerability in the Model Context Protocol PHP SDK (mcp/sdk) versions 0.5.0 through 0.7.0. The HTTP client transport reads Server-Sent Events (SSE) streams into an unbounded in-memory buffer, which only flushes upon receiving a double newline delimiter. A malicious or compromised server that never sends this delimiter can cause the buffer to grow indefinitely, exhausting client memory and causing a denial of service. This affects any client using HttpTransport against an untrusted server and requires no authentication or user interaction beyond connection initiation. The issue is fixed in version 0.7.1.
CVE-2026-55637 is an origin validation error in the Streamable HTTP transport of the genieacs-mcp server prior to version 0.3.2. The vulnerability arises because the /mcp listener on 127.0.0.1:8080 does not validate Host or Origin headers when MCP_AUTH_TOKEN is unset, allowing DNS rebinding attacks. Exploitation can lead to unauthorized initialization of MCP sessions and invocation of operations against the GenieACS NBI, potentially exposing or modifying CPE management state such as device reboots and firmware tasks. The issue is fixed in version 0.3.2.
CVE-2026-55557 is a path traversal vulnerability in That1Drifter's browse-mcp server prior to version 0.8.2. The flaw allows an attacker to write arbitrary files to the host filesystem by controlling the save directory and file contents. This can lead to overwriting critical files such as ~/.bashrc, autostart entries, or cron files, potentially enabling code execution on the host. The vulnerability arises from insufficient validation of caller-controlled paths and a fetch fallback that bypasses origin restrictions. The issue is fixed in version 0.8.2.
Nextcloud MCP Server versions prior to 0.117.2 have a missing authentication vulnerability on the POST /webhooks/nextcloud endpoint. The default configuration allows unauthenticated requests because the WEBHOOK_SECRET is unset by default and not validated at startup. This enables an attacker to send forged webhook payloads that specify arbitrary user IDs, leading to unauthorized deletion or re-indexing of vector embeddings and destruction of the semantic search index. The issue is fixed in version 0.117.2.
CVE-2026-55529 is a vulnerability in MervinPraison's PraisonAI multi-agent teams system prior to version 4.6.58. The issue arises because the MCP HTTP Stream _validate_origin method incorrectly validates origins by using a startswith check, allowing attacker-controlled origins like localhost.evil.example to bypass the localhost allowlist. This enables a malicious webpage to submit requests to the local MCP server without an API key and execute exposed tools. The vulnerability is fixed in version 4.6.58.
CVE-2026-53509 is a server-side request forgery (SSRF) vulnerability in ondata's ckan-mcp-server versions prior to 0.4.106. The vulnerability arises because the server validates caller-supplied CKAN server URLs by inspecting only the parsed hostname string, missing certain hostname aliases like 'ip6-localhost' that resolve to loopback addresses. This allows remote attackers who can invoke CKAN tools with a server_url parameter to make the server connect to local or private network addresses and potentially receive data from those internal endpoints. The issue was fixed in version 0.4.106 by expanding the blocked hostname checks to include these aliases.
CVE-2026-59279 is a high-severity vulnerability in Spring AI version 2.0.0 where the MCP Streamable HTTP server transport (WebFlux and WebMvc variants) does not limit the number of sessions retained and does not require client authentication by default. This allows a remote attacker to cause the server to accumulate an unlimited number of sessions, leading to memory exhaustion and denial of service for legitimate clients.
CVE-2026-18482 is a critical OS command injection vulnerability in Klarso GmbH's neo-mjs product. The flaw exists in the FileSystemService.mjs component of the ai/mcp/server/file-system MCP server, where the functions checkSyntax() and runPlaywrightTest() unsafely interpolate user-controlled absolutePath values into shell commands. This allows an attacker to execute arbitrary operating system commands if an AI agent is tricked into invoking these functions. The vulnerability is fixed in commit 88c77fc.
CVE-2026-75149 is a high-severity code injection vulnerability in marimo versions before 0.23.15. It affects the notebook configuration handler, allowing attackers to execute arbitrary local commands by embedding a crafted MCP server entry in a notebook. This command is executed as a subprocess immediately when the notebook is opened in edit mode, without requiring authentication or execution of any notebook cells. The vulnerability is fixed in version 0.23.15.
The npm package 'broadcast-graphics-mcp' version 1.0.0 contains malicious code in its postinstall script that automatically executes upon installation. This script collects host identifiers such as hostname, platform, architecture, Node.js version, package name, npm lifecycle event, and timestamp, then sends this data as JSON to a hardcoded external server controlled by the package author. This behavior results in the exfiltration of installer machine metadata without user consent or configuration options.
Context7 versions up to 2.1.2 have a prompt injection vulnerability in the Custom AI Instructions feature. This flaw allows attackers to inject malicious instructions into connected AI coding agents via the MCP server. Exploitation can lead to credential exfiltration from environment files and destructive file deletion triggered by routine agent actions.
CVE-2026-50143 is a Server-Side Request Forgery (SSRF) vulnerability in the apify-mcp-server prior to version 0.10.11. It allows a malicious Actor publisher to manipulate URLs to redirect requests to third-party hosts, exposing the victim's Apify API token and enabling unauthorized access to Actors, stored data, and billable compute. The vulnerability is fixed in version 0.10.11.