Quick answers
- Is CVE-2026-111768 exploited in the wild?
- No. Junglewise Threat Intelligence has no report of CVE-2026-111768 being exploited in the wild.
- How severe is CVE-2026-111768?
- High severity, with a CVSS score of 7.5, published 5 November 2026.
- Which products and versions are affected?
- Anthropic @Modelcontextprotocol/Sdk 1.3.0 through 1.31.0; Anthropic @Modelcontextprotocol/Server 2.0.0 through 2.2.0; @modelcontextprotocol/sdk (npm) >= 1.3.0, < 1.32.0, SEMVER:introduced=1.3.0,fixed=1.32.0; @modelcontextprotocol/server (npm) >= 2.0.0, < 2.3.0, SEMVER:introduced=2.0.0,fixed=2.3.0.
- Is there a fix for CVE-2026-111768?
- Yes. @modelcontextprotocol/sdk (npm): upgrade to 1.32.0 or later; @modelcontextprotocol/server (npm): upgrade to 2.3.0 or later. Until then: Set the new `maxToolInputElements` option to a reasonable limit just above your largest expected input (e.g., `new McpServer(info, { maxToolInputElements: 10_000 })`), and lower the request size limit via `maxRequestBodySize` on the transport or parser to the minimum your tools require. If upgrading is not yet possible, reducing `maxRequestBodySize` alone will help; versions 1.30.1 and 2.1.0+ support this option. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.
- How is CVE-2026-111768 attacked?
- It is attacked over the network, in a software library.
- Am I affected by CVE-2026-111768?
- Check whether you run Anthropic @Modelcontextprotocol/Sdk, Anthropic @Modelcontextprotocol/Server, @modelcontextprotocol/sdk (npm) or one of the others listed at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.
How to fix
- @modelcontextprotocol/sdk (npm): upgrade to 1.32.0 or later
npm install @modelcontextprotocol/sdk@1.32.0 - @modelcontextprotocol/server (npm): upgrade to 2.3.0 or later
npm install @modelcontextprotocol/server@2.3.0
Set the new `maxToolInputElements` option to a reasonable limit just above your largest expected input (e.g., `new McpServer(info, { maxToolInputElements: 10_000 })`), and lower the request size limit via `maxRequestBodySize` on the transport or parser to the minimum your tools require. If upgrading is not yet possible, reducing `maxRequestBodySize` alone will help; versions 1.30.1 and 2.1.0+ support this option.
Technical details
The vulnerability is a resource exhaustion flaw in request validation: servers accept unbounded numbers of array elements and object keys during validation, allowing a single HTTP request under the 4 MiB default size limit to consume arbitrary memory and trigger a JavaScript heap out-of-memory abort. The attack requires network access to an HTTP-exposed McpServer with tools that accept arrays or records as input; stdio-only servers are not vulnerable. The advisory provides two mitigations: setting the new `maxToolInputElements` option and lowering `maxRequestBodySize` to the minimum needed.
Affected products
- Anthropic @modelcontextprotocol/sdk 1.3.0 through 1.31.0
- Anthropic @modelcontextprotocol/server 2.0.0 through 2.2.0
Identifiers
- CVE-2026-111768
- GHSA-46jr-4wx8-v453
Timeline
- 2026-10-20: disclosed