CVE-2026-44895
GitLab MCP Server lets an AI agent talk directly to GitLab. Prior to 0.6.0, the HTTP transport in src/transport.ts ships with no authenticat
CVSS
—
Sin CVSS
EPSS
0.4%
p33
KEV
—
Exploit Today
10
0-100
Publicado: 26 may 2026 · Última mod.: 23 jul 2026 · CWE-306 · CWE-942
0.4%EPSS · 30 días0.4%
2026-08-242026-09-22
GitLab MCP Server lets an AI agent talk directly to GitLab. Prior to 0.6.0, the HTTP transport in src/transport.ts ships with no authentication layer at all and a wildcard Access-Control-Allow-Origin: * on every response. The structural defect is that the SSE server stands up a stateful, mutation-capable RPC endpoint that is backed by the operator's GITLAB_PERSONAL_ACCESS_TOKEN without any inbound credential check, then advertises itself to every cross-origin browser context via the wildcard CORS header. The httpServer.listen(port) call at line 97 also passes no host argument, so the bind defaults to 0.0.0.0 and exposes the auth-less surface on every interface. This vulnerability is fixed in 0.6.0.
CVECVSSEPSSKEVRExplotTítuloVis.
CVE-2026-772549.1 CRÍ—
——0MCP Atlassian is a Model Context Protocol (MCP) server for Atlassian products (Confluence and Jira). Prior to 0.22.0, requests to the HTTP MCP endpoint without a per-user identity are allowed to reach tool handlers, which then use globally configured Jira or Confluence credentials. A network caller can perform operations with the operator account's permissions unless the deployment has an independent authentication boundary. The advisory traces the vulnerable input and processing flow through streamable-http, UserTokenMiddleware, _get_fetcher, and global credentials, which identify the affected entry points, controls, and code paths. This issue is fixed in version 0.22.0.43mCVE-2026-772488.6 ALT—
——0MCP Atlassian is a Model Context Protocol (MCP) server for Atlassian products (Confluence and Jira). Prior to 0.22.0, the streamable HTTP transport accepts requests without a user identity and falls back to operator credentials, while upload_attachment accepts an unrestricted file_path. An unauthenticated network caller can read files available to the MCP process, upload them to an attacker-selected Jira issue or Confluence page, and retrieve the contents. The advisory traces the vulnerable input and processing flow through streamable-http, UserTokenMiddleware, upload_attachment, file_path, and _get_fetcher, which identify the affected entry points, controls, and code paths. This issue is fixed in version 0.22.0.43mCVE-2026-944557.1 ALT—
——0An HTTP endpoint intended for provisioning enterprise and reseller organisations is reachable without any session. The authentication middleware is bound only to an explicit list of controllers, and the enterprise controller is not on that list, so no authentication runs for these routes.
The endpoint's only check is that the request body carries a token bearing a valid signature from the instance secret. It does not check what that token was issued for. Login tokens are signed with the same secret and carry no purpose, audience or expiry claim, so an ordinary user's own session token satisfies the check.
Presented with such a token, the endpoint creates a new organisation holding the highest subscription tier, flagged as lifetime and with a channel allowance far above any sold plan, creates an organisation-owner account alongside it, and returns the new organisation's API key in the response body. That key is immediately valid against the public API.23mCVE-2026-651148.3 ALT—
——0NVIDIA Infrastructure Controller for Linux contains a vulnerability where an attacker could cause missing authentication for a critical function. A successful exploit of this vulnerability might lead to data tampering, denial of service, and information disclosure.23mCVE-2026-757918.6 ALT—
——0Zohocorp ManageEngine ADSelfService Plus versions before build 7001 are vulnerable to an authentication bypass vulnerability in the REST API.28mCVE-2026-90882——
——0The open-vsx.org deployment returned Access-Control-Allow-Origin reflecting the requesting origin together with Access-Control-Allow-Credentials: true on the authenticated /user/ endpoints. A page on any origin could therefore issue credentialed requests to the service in a logged-in user's browser and read the responses.
This exposed /user (login name, avatar, homepage, tokens URL), /user/tokens, /user/namespaces, /user/extensions, /user/search/{name} and /user/namespace/{name}/members, and — because /user/csrf was readable the same way — allowed the CSRF protection on write endpoints to be defeated. Chaining the two, an attacker page could call /user/token/create and exfiltrate a personal access token carrying publish and delete rights over the victim's namespaces.
The headers were emitted by the CDN/edge layer, not by the application: the Open VSX software sets allowCredentials(true) in exactly one place, against a single exact origin derived from ovsx.webui.url, and defines no CORS mapping on /user/ beyond it. No configuration of the software produces origin reflection with credentials.50m