CVE-2026-108542 is a server-side request forgery (SSRF) vulnerability in the 021is elvix-sdk affecting versions 0.10.0 and 0.10.1. The flaw exists in the MCP Request Handler component, specifically in the src/mcp/index.ts file, where manipulation of the path argument allows remote attackers to induce SSRF. The vulnerability has a medium severity with a CVSS 4.0 score of 5.3. Exploit code has been publicly disclosed, but no known active exploitation in the wild has been reported. The vendor was contacted but did not respond, and no patch or remediation is currently available.
1MCP Agent versions 0.20.0 through 0.39.0 have an incorrect authorization vulnerability that allows authenticated clients to bypass OAuth tag-scope enforcement. By using negated advanced tag-filter expressions, attackers with a single-tag token can access backend MCP server tools beyond their authorized scopes. This vulnerability has a medium severity rating with a CVSS score of 5.3.
zotero-mcp versions 0.10.0 through 0.14.1 contain a server-side request forgery (SSRF) vulnerability. This flaw allows attackers to manipulate the _fetch_embedded_metadata function to fetch URLs without validating the destination, potentially accessing internal services. Attackers can use prompt injection to trigger zotero_add_by_url, causing requests to internal or link-local hosts, which may leak citation metadata and error information. The vulnerability has a low severity score and no known exploits in the wild.
GenOffice versions up to 0.11.505 have an incorrect permission assignment vulnerability in the HTTP MCP server file store. This flaw allows local unprivileged users to read uploaded and generated documents by accessing a world-readable directory under the system temporary folder. The vulnerability bypasses HTTP bearer token protections, exposing client uploads and converted outputs to unauthorized local users.
x64dbg-MCP Server versions prior to 1.1 expose the debugger's full functionality over HTTP without any authentication, allowing unauthenticated network clients to execute arbitrary debugger commands, attach to processes, and manipulate memory and files. This critical vulnerability is fixed in version 1.1.
CVE-2026-107820 is an integer overflow vulnerability in the duty1g x64dbg-MCP Server plugin prior to version 1.2. The vulnerability arises from parsing an unbounded Content-Length header and performing unchecked addition, which can cause a runtime integer-overflow panic. This leads to termination of the x64dbg process and its live debugging session, resulting in a denial of service. The server listens on all interfaces by default, allowing unauthenticated network clients to trigger this condition. The issue is fixed in version 1.2.
CVE-2026-104634 is an incorrect type conversion vulnerability in the BeamMCP.Server component of ScriptKittyOS beam_mcp. It causes boolean values true, false, and null from an MCP client to be converted into strings "true", "false", and "nil" when passed to the host's dispatch function. This leads to logic errors where boolean checks behave incorrectly, potentially causing false to be treated as true. The issue affects beam_mcp versions from 0.1.0 up to but not including 0.10.1.
Improper Input Validation vulnerability in BeamMCP.Schema in ScriptKittyOS beam_mcp allows an MCP client to reach a tool's dispatch function with arguments that violate the input schema the server advertised. BeamMCP.Schema.validate/2 checked type, required, additionalProperties, enum and numeric bounds on the top-level arguments object only. Constraints inside nested objects and on array items (items, minItems, maxItems, minLength, maxLength, pattern, nested required, enum and additionalProperties: false) were advertised by tools/list and never checked at tools/call or prompts/get, and keywords outside the enforced subset (oneOf, anyOf, $ref) were advertised and ignored.
A host whose dispatch code relies on the schema it declared receives values the schema forbids, such as an out-of-range number or an undeclared key inside a nested object. What the host does with such a value decides the impact.
This issue affects beam_mcp: from 0.1.0 before 0.10.1.
Critical and high-severity vulnerabilities were identified in SonicWall SMA1000 appliances and Splunk products, including Splunk Enterprise and MCP Server. The SonicWall vulnerabilities include a critical pre-authenticated SSRF flaw (CVE-2026-102255) that could allow unauthenticated attackers to perform unauthorized internal requests. Additional vulnerabilities in SonicWall could lead to remote code execution and cross-site scripting. Splunk patched multiple critical bugs enabling arbitrary command execution, unauthorized access, and code injection. No evidence of exploitation in the wild has been reported. SonicWall SSL-VPN products are not affected by these vulnerabilities.
CVE-2026-76286 affects Splunk MCP Server versions 1.2 up to but not including 1.2.1. The vulnerability allows a user with the mcp_tool_execute capability to run a custom API tool that sends the user's Splunk platform authentication token to a URL controlled by another user. This could enable the attacker to capture the token and impersonate the original user to access data and perform actions. The issue arises because the server does not sufficiently verify that the URL request is sent to the expected destination.
### Summary
In affected versions, the SDK's OAuth client let the MCP server decide which authorization server received the client's OAuth credentials. Credentials were not tied to the authorization server they belong to.
A malicious or compromised MCP server could name its own authorization server. With no user interaction, the client would send it:
- the `refresh_token` and `client_secret` stored from an earlier sign-in
- the `client_secret` or signed assertion configured on a bundled provider
### Am I affected?
Yes, if both of these hold:
- your application uses the SDK's OAuth client over HTTP, through any of:
- an `authProvider` on a transport
- the `withOAuth()` middleware
- direct calls to `auth()` or `fetchToken()`
- it may connect to an MCP server you do not fully trust while holding credentials for a legitimate authorization server
Affected versions:
- `@modelcontextprotocol/sdk` 1.12.0 through 1.30.1
- `@modelcontextprotocol/client` 2.0.0 through 2.1.0, only for:
- bundled providers without `expectedIssuer`
- credentials stored or supplied without `issuer`
- direct calls to `fetchToken()`
- providers that read storage back through `OAuthTokensSchema` or `OAuthClientInformationSchema`
Not affected:
- MCP servers built with the SDK
- stdio clients
### Fix
Upgrade to:
- 1.x: `@modelcontextprotocol/sdk` 1.31.0 or later
- 2.x: `@modelcontextprotocol/client` 2.2.0 or later, and `@modelcontextprotocol/core` 2.2.0 or later if you import it directly
The client now records the authorization server as `issuer` on saved credentials and does not send them to a different one. The user signs in again, or the call throws.
In the following cases, upgrading to the patched version is not enough. You also need to make a change:
- **Bundled providers** (`ClientCredentialsProvider`, `PrivateKeyJwtProvider`, `StaticPrivateKeyJwtProvider`, `CrossAppAccessProvider`): pass `expectedIssuer`, for example `expectedIssuer: 'https://auth.example.com'`. Without it they still use whichever authorization server the MCP server names.
- **Credentials saved without `issuer`:** tokens and client information your `OAuthClientProvider` persisted (file, keychain, database) before upgrading, including everything 1.x saved before 1.31.0. They still go to whichever authorization server is named at first use. Add `issuer` to them, or clear them so users sign in again.
- **Your own `OAuthClientProvider`:** save exactly what `saveTokens()` and `saveClientInformation()` are given, including `issuer`. For pre-registered credentials, include `issuer` in what `clientInformation()` returns.
Not covered by this fix:
- a new interactive sign-in, which still goes to the authorization server the MCP server names. Only complete sign-ins for servers you trust.
- `refreshAuthorization()` and `exchangeAuthorization()` called directly
- 2.x with `skipIssuerMetadataValidation: true`
If an affected client may have connected to an untrusted MCP server, rotate its client secret or signing key and revoke its tokens.
If you cannot upgrade yet, connect OAuth clients only to MCP servers you trust. 2.0.0 and 2.1.0 already accept `expectedIssuer`.
Microsoft UFO is an open-source framework for intelligent automation across devices and platforms. Prior to 3.0.9, the execute_command tool in ufo/client/mcp/http_servers/linux_mcp_server.py treats sort and uniq as read-only commands while the free-form command parameter can select their file-output forms. An authenticated caller can use sort -o or the optional second uniq operand to create or overwrite files writable by the UFO server process without shell metacharacters, because the allowed binary opens the destination itself and the argument policy does not reject the operation. This can corrupt configuration or other writable data and disrupt the service, but the demonstrated primitive does not directly disclose files or establish arbitrary code execution. This issue is fixed in version 3.0.9.
Langflow is a tool for building and deploying AI-powered agents and workflows. From 1.5.0 until 1.10.3, an IP spoofing vulnerability in the Model Context Protocol (MCP) configuration installation endpoint (POST /api/v1/mcp/project/{project_id}/install) allowed authenticated remote attackers to bypass the "local-only" access restriction. By sending a spoofed X-Forwarded-For: 127.0.0.1 header, an attacker could make the server treat the request as originating from localhost, letting them write/overwrite an MCP client configuration file on the server's filesystem. This vulnerability is fixed in 1.10.3.
Langflow is a tool for building and deploying AI-powered agents and workflows. Prior to 1.9.0, any authenticated Langflow user can achieve Remote Code Execution (RCE) on the server by adding an MCP server with the "Stdio" transport. The user-supplied command field is passed directly to bash -c "exec {command}" with zero validation, no allowlisting, and no sandboxing. The command executes immediately when the server list is fetched. Additionally, the env field allows arbitrary environment variable injection (e.g., LD_PRELOAD, PATH override). This vulnerability is fixed in 1.9.0.
MCP Atlassian versions prior to 0.22.0 contain a server-side request forgery (SSRF) protection bypass vulnerability. The issue arises from the validate_url_for_ssrf function misinterpreting backslash characters in the authority component of URLs, leading to authority confusion. This allows crafted URLs to appear as external hosts during validation while the HTTP client connects to internal hosts, potentially accessing protected network resources. The vulnerability is fixed in version 0.22.0.
An SSRF vulnerability exists in mcp-atlassian versions from 0.11.9 up to but not including 0.22.0. This flaw allows an attacker who can set specific HTTP headers to cause the server to send requests to internal or metadata-service URLs, bypassing prior incomplete SSRF mitigations. The issue is resolved in version 0.22.0.
MCP Atlassian versions prior to 0.22.0 have a Server-Side Request Forgery (SSRF) vulnerability due to missing redirect protection in sessions created via basic-auth and OAuth authentication branches. This allows attacker-controlled or compromised Atlassian instances to cause the server to follow redirects to internal addresses without proper validation. The issue is fixed in version 0.22.0.
MCP Atlassian versions prior to 0.22.0 contain a vulnerability in their HTTP upload tools that allows remote attackers with tool access to specify arbitrary server-local file paths. This flaw enables reading sensitive local files and uploading them as attachments to Atlassian products like Jira and Confluence. The issue is resolved in version 0.22.0.
A vulnerability in MCP Atlassian prior to version 0.22.0 allows an attacker to read arbitrary local files on the server and attach them to a Jira issue. This occurs because the update_issue attachments argument is converted into local paths without proper workspace validation, enabling a confused deputy attack. The issue is fixed in version 0.22.0.