Back Gbhackers AWS Warns of MCP Gateway Registry Vulnerability Allowing Attackers to Access Internal Network Services
AWS has disclosed a server-side request forgery (SSRF) flaw in mcp-gateway-registry, an open-source registry and gateway for Model Context Protocol (MCP) servers.
Tracked as CVE-2026-125926 under bulletin 2026-180-AWS, the issue lets an authenticated user make the gateway send requests to services on a deployment’s internal network. AWS rates the bulletin “Important”, and the GitHub advisory scores it 8.5 (High) under CWE-918.
AWS MCP Gateway Registry Vulnerability
The registry centralizes discovery, access control, and proxying of registered MCP tool assets. A vulnerability exists in the server registration and update API, which failed to validate the upstream address of a registered server. A user with permission to register or modify servers could potentially redirect that address to an internal service.
As a result, the gateway might send requests to this service and, in some instances, return responses directly to the user. This flaw affects versions from v1.0.3 through v1.34.0.
The risk grows in cloud deployments. Depending on network configuration, reachable targets can include the cloud instance metadata service and the credentials it issues.
Attackers already target this path. Researchers recently observed MCP reconnaissance paired with cloud metadata SSRF to steal credentials. A registry running with an attached cloud role could therefore expose role credentials, letting an attacker move from application-layer abuse to wider cloud-account access.
The flaw also fits a broader pattern in AI agent infrastructure. AWS recently fixed similar SSRF and authentication flaws in its Loom agent platform , where users with write scopes could redirect backend connections to internal endpoints. Registration APIs that accept URLs are attractive targets because they are trusted by design and are often reachable by many authenticated users.
AWS fixed the issue in version 1.35.0, which validates upstream addresses when a server is registered or modified. Administrators should also note a version-sequencing trap. A separate management API issue was fixed in 1.25.0.
However, deployments on 1.25.0 or later stay vulnerable to this flaw until they reach 1.35.0. The 1.35.0 release notes cover the upgrade steps. These include a breaking change for Helm installs with bundled Keycloak, which must adopt the keycloak-client-secret Secret before upgrading.
The upgrade does not clean up earlier configurations. Server entries stored before the upgrade stay unchanged, so customers who ran an affected version should take these steps:
Review the upstream addresses of registered servers, and correct or remove any address they did not intend.
Review application logs from the affected period for unexpected server registrations and changes.
If the registry runs with an attached cloud role and the logs cannot rule out unexpected activity, review that role’s activity and rotate its credentials.
Teams that cannot upgrade yet can reduce exposure. They can limit registration and modification rights to trusted accounts and restrict outbound network access from the registry and auth-server components.
On Amazon EC2, they should require IMDSv2 and grant each component’s role only the permissions it needs. IMDS abuse for cloud initial access shows why IMDSv2 matters, since IMDSv1 accepts unauthenticated requests that SSRF can exploit. These measures do not replace the post-upgrade actions.
AWS credited external researcher novice-22 for coordinated disclosure. With thousands of MCP servers already found vulnerable, defenders should treat MCP registries as sensitive infrastructure. Questions can go to [email protected] .
Give your SOC the alert context on the latest threats to stop incidents quickly -> Integrate Threat Intel from 16K SOCs :
Artificial Intelligence
Cyber security Course
Cyber Security Resources
Cybersecurity
Information Gathering
Information Security Risks
Massive Microsoft 365 Outage Disrupts Outlook, Teams, Copilot, and Defender XDR Worldwide
nDiligence Releases Four-Year Study Identifying a Critical Supply-Chain Vulnerability Across 50 Platform Types
CISA Warns of Actively Exploited Microsoft Outlook Zero-Day Vulnerability
ESET Blocks Over 700,000 Cruciferra Malware Attacks Using Advanced Encryption to Evade Antivirus Detection
Global Scam Networks Use AI, Deepfakes, and Crypto to Steal Billions From Victims, FATF Warns
Critical MongoDB Use-After-Free Vulnerability Allows Remote Code Execution Without Authentication
The full story
This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.
