EUVD-2026-89360
GHSA-4qvr-m43m-5qgr
Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessReach exposed /mcp-proxy/server/stdio endpoint
-
DeliverySend crafted GET with STDIO transport parameters
-
ExploitcreateTransport spawns attacker-controlled command
-
ExecutionExecute arbitrary code on host
-
ImpactFull host compromise
Vulnerability Assessment
| Weakness |
CWE
OWASP Top 10:2025
A05:2025 - Injection
ASVS 5.0
V1 - Encoding and Sanitization Code injection prevention, sandbox exec
MITRE ATT&CK
2 techniques across 2 tactics: Execution, Initial Access The technique list with ATT&CK links and the kill-chain view is part of the Basic plan and up. |
| Exploitation Conditions | Exploitation requires network reachability to MetaMCP's internal MCP inspector proxy endpoint GET /mcp-proxy/server/stdio (the STDIO branch of createTransport in routers/mcp-proxy/server.ts) on a version at or below 2.4.22. The CVSS vector (PR:N/UI:N) indicates no authentication or user interaction is needed against that endpoint, and SSVC marks it automatable. The key limiting factor is exposure: this is a developer/inspector proxy interface, so realistic risk hinges on whether it is bound to a network-accessible interface rather than localhost and whether any authentication or reverse-proxy control fronts it in the specific deployment. Where the endpoint is exposed, a single crafted request yields code execution (Technical Impact: total). |
| Impact | The CVSS 3.1 base score of 9.8 (AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H) describes an ideal-case unauthenticated remote code-execution flaw, and CWE-94 (code injection) plus the RCE/Code Injection tags are consistent with the described defect: the STDIO branch of createTransport in the /mcp-proxy/server/stdio inspector proxy endpoint spawns a local process from request-controlled input. SSVC reinforces the impact side - Technical Impact: total and Automatable: yes - meaning a working exploit would be reliable and scriptable. However, the two exploitation signals point the other way: SSVC Exploitation is 'none' and there is no KEV listing and no public exploit identified at time of analysis, so this is a high-severity but not-yet-exploited issue rather than an active in-the-wild threat. The most important caveat is reachability: MetaMCP's MCP inspector proxy is an internal/developer-facing endpoint, and the PR:N/AV:N rating assumes it is network-exposed without authentication. If the endpoint is bound to localhost or sits behind the app's auth layer in a given deployment, real-world risk drops well below the 9.8 headline - verify exposure with the vendor advisory. Note also that NVD/EUVD carry no usable CPE ('cpe:2.3:a:n/a:n/a:*' is a placeholder) and affected versions are recorded only as 'n/a', so version scoping relies on the description's 'up to and including 2.4.22'. |
| Exploit ScenarioDerived | An attacker who can reach the MetaMCP MCP inspector proxy sends a crafted GET request to /mcp-proxy/server/stdio whose parameters drive the STDIO transport's createTransport code path, causing the server to spawn an attacker-specified command/arguments and execute arbitrary code on the host. Per CVSS this needs no authentication, no user interaction, and low complexity, and SSVC marks it automatable; no public exploit code has been identified yet, so exploitation currently requires the attacker to craft the request from the source in routers/mcp-proxy/server.ts. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Sep 29, 2026
Unauthenticated remote code execution affects metatool-ai MetaMCP through version 2.4.22 via the internal MCP inspector proxy endpoint GET /mcp-proxy/server/stdio, where request-controlled input reaches process spawning in the STDIO branch of createTransport (routers/mcp-proxy/server.ts). The CVSS 3.1 base score of 9.8 (AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H) indicates a network-reachable, unauthenticated attacker can achieve total confidentiality, integrity, and availability impact with no user interaction, and CISA's SSVC assessment rates the flaw automatable with total technical impact. No confirmed active exploitation (CISA KEV) and no public exploit code were identified at time of analysis; the practical risk depends primarily on whether the inspector proxy endpoint is exposed beyond localhost in a given deployment.
RemediationAI
Patch status: no vendor-released patch or fixed version is identified in the available intelligence, so remediation must be driven by the vendor repository (https://github.com/metatool-ai/metamcp) and the advisory at https://www.traceforce.ai/security-advisories/cve-2026-79538 - confirm whether a release above 2.4.22 is published and upgrade to it as soon as one is available. Until then, the highest-value compensating control is to remove network exposure of the inspector proxy: bind MetaMCP's /mcp-proxy/ routes to 127.0.0.1 only and never publish that port, since the attacker path requires network reachability to GET /mcp-proxy/server/stdio (trade-off: developers lose remote access to the inspector and must tunnel or use a local instance). If remote access is unavoidable, place the endpoint behind an authenticating reverse proxy or mTLS gateway and restrict it by source IP to trusted developer networks (trade-off: adds an auth layer that must be kept in sync with MetaMCP's own routing and may still allow lateral abuse by any authenticated user). Where the inspector proxy is not needed in production, disable or blocklist the /mcp-proxy/* route entirely at the reverse proxy or network policy layer (trade-off: any legitimate inspector or debugging workflows break). Because SSVC rates the flaw automatable with total impact, treat any host that has run an exposed MetaMCP instance as potentially compromised, rotate credentials and secrets accessible to the MetaMCP service account, and audit for unexpected child processes or outbound connections; running MetaMCP as a low-privilege, containerized user with a minimal filesystem limits the blast radius of a successful spawn but does not prevent the code execution itself.
Affected VersionsENISA
Affects your stack
Find out whether the stacks and servers you run are on an affected version.
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
The defect is a CWE-94 code injection in MetaMCP, an open-source Model Context Protocol (MCP) gateway/aggregator written for Node.js. MetaMCP includes an internal MCP inspector proxy used for developer debugging of MCP server connections; its createTransport function in routers/mcp-proxy/server.ts contains an STDIO branch that launches a local child process, and the parameters for that launch are derived from the incoming HTTP request to GET /mcp-proxy/server/stdio rather than from a fixed, server-side allowlist. Because attacker-controlled data flows into a process-spawn sink, a single crafted request can execute arbitrary commands or code under the privileges of the MetaMCP service account. The CVSS vector's PR:N/UI:N components mean the endpoint itself enforces no authentication and requires no victim interaction, and the SSVC 'Automatable: yes' rating means exploitation can be scripted at scale once a target endpoint is reachable. The provided CPE string is a placeholder (n/a:n/a) and the EUVD affected-version list is also unspecified, so version identification must rely on the vendor repository and the Traceforce advisory referenced in the CVE record; the affected range is stated as MetaMCP up to and including 2.4.22.
Know when N A gets its next vulnerability
- Watch products and get an email when something new is published
- Check this CVE against your own stack and servers
- Exploit and PoC references, threat intelligence from 22 sources
No credit card · 7-day full trial
Severity by source
AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
STDIO branch spawns attacker-controlled processes from unauthenticated requests to the proxy endpoint (PR:N, AC:L), yielding full host code execution (C/I/A:H); AV:N assumes the internal endpoint is network-exposed.
Primary rating from Vendor (mitre).
CVSS VectorVendor: mitre
Lifecycle Timeline
3DescriptionCVE.org
metatool-ai MetaMCP up to and including 2.4.22 is vulnerable to Code Execution in the internal MCP inspector proxy endpoint GET /mcp-proxy/server/stdio (createTransport, STDIO branch, routers/mcp-proxy/server.ts).
Executive ViewAI
metatool-ai MetaMCP through version 2.4.22 contains a critical unauthenticated remote code execution vulnerability that could allow an attacker to fully compromise affected systems. Organizations running this AI tooling are at risk if the inspector proxy endpoint is exposed beyond localhost. Leadership should prioritize immediate exposure assessment and mitigation because no vendor patch is yet available.
No vendor-released patch identified at time of analysis. Interim mitigations include: restrict the /mcp-proxy/server/stdio endpoint to localhost or trusted networks only, disable the MCP inspector proxy if not required, implement network segmentation and authentication via a reverse proxy, and monitor logs for unexpected child process creation or outbound connections from MetaMCP servers.
Threat Intelligence Overview
NVD
CVE.org
EPSS
Vulnrichment
How this decision was reached
The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.
What every action level means
SSVC AssessmentENISA
Priority Score
Attack Technique
Code injection occurs when an application accepts user input and passes it directly into a language interpreter or evaluator without proper sanitization.
Remote Code Execution represents the critical moment when an attacker successfully runs arbitrary code on a target system without physical access.
LibreNMS before 24.10.0 allows a remote attacker to execute arbitrary code via OS command injection involving AboutContr
Reachable assertion in Open5GS up to version 2.7.6 allows remote unauthenticated attackers to crash the Diameter CER han
A Command Injection issue in the payload build page in BYOB (Build Your Own Botnet) 2.0 allows attackers to execute arbi
Denial of service in Open5GS 2.7.6 and earlier allows network-adjacent attackers with low-privilege Diameter Gx peer acc
Denial of service in Open5GS HSS (Home Subscriber Server) versions 2.7.0 through 2.7.6 is triggered remotely by manipula
Remote command injection in Control Web Panel allows unauthenticated attackers to execute arbitrary OS commands as root
LINQPad before 5.52.01 Pro edition is vulnerable to Unsafe Deserialization in LINQPad.AutoRefManager::PopulateFromCache(
Remote code execution in dash-uploader (Python package for Plotly Dash) versions 0.1.0 through 0.7.0a2 allows unauthenti
Command injection in LiteLLM 1.18.10 lets attackers who can configure MCP (Model Context Protocol) servers supply arbitr
SQL injection in AzuraCast's backup restore flow (versions up to and including 0.23.7) lets an attacker achieve privileg
Arbitrary file read in Yu AI Code Mother v4.3 lets anonymous, remote attackers retrieve files outside the intended previ
Remote code execution in APScheduler (all versions through 3.10.x and 4.0.0a5) is achievable when applications deseriali
Exploit-DB