← All advisories
CVE-2026-54032mediumAgent hijacking2026-11-20

LibreChat: a malicious remote MCP server could run script via its description, MCP OAuth followed redirects to internal URLs, and another user's file could be pulled into the model context, plus 7 more CVEs (CVE-2026-54023 to 54046; fixed 0.8.6–0.8.9)

Alongside the two LibreChat CVEs covered earlier, NVD published ten more on 20 November 2026 (GitHub advisories from 17 November). The AI-specific ones: MCPConfigDialog rendered a remote MCP server's description through dangerouslySetInnerHTML, so a malicious MCP server could run script in the LibreChat origin when a user opened its config dialog (CVE-2026-54032, fixed 0.8.6); MCP OAuth discovery, registration, token, refresh and revocation requests validated only the first URL and then followed redirects to internal addresses (CVE-2026-54031, fixed 0.8.8); and chat attachments accepted another user's file_id, putting that user's extracted document text into the attacker's model context (CVE-2026-54043, fixed 0.8.7). 0.8.5 also sent federated OAuth access, ID and refresh tokens to the browser via GET /api/user (CVE-2026-54023). Upgrade to 0.8.9 to get all fixes.

01.

What it is

CVE-2026-54032 (GHSA-vfg3-f8h3-2683, CVSS 6.1): client/src/components/Chat/Input/MCPConfigDialog.tsx; a secondary sink in Banners/Banner.tsx rendered admin banner HTML without DOMPurify. Fixed 0.8.6.

CVE-2026-54031 (GHSA-f7xq-83jc-cgrj, CVSS 6.3): createOAuthFetch in packages/api/src/mcp/oauth/handler.ts. The MCP transport path already validated redirects and is not affected. Fixed 0.8.8.

CVE-2026-54043 (GHSA-wwj3-x5mw-wq7x, CVSS 6.3): updateFileUsage in packages/data-schemas/src/methods/file.ts did not scope file_id lookups to the owner; file ids can leak through GET /api/share/{shareId}. It can also strip the victim file's expiry. Fixed 0.8.7.

CVE-2026-54023 (GHSA-p84h-536v-gm3g, CVSS 6.5): getUserController serialised federatedTokens and tokenset in 0.8.5 only. Fixed 0.8.6.

Also fixed: cross-user message injection via body conversationId (CVE-2026-54026, 0.8.6); plaintext password logged on login validation errors (CVE-2026-54041, low, 0.8.6); no attempt limit on 2FA verify-temp (CVE-2026-54042, 0.8.7); password reset wiping email-verification tokens (CVE-2026-54045, 0.8.7); unauthenticated account enumeration via /api/user/verify (CVE-2026-54046, 0.8.7); ReDoS in principal search (CVE-2026-54038, 0.8.9).

02.

How to check

  1. Check your LibreChat version (release tag or container image tag).
  2. List remote MCP servers configured by users or admins and remove any you don't trust.
  3. On 0.8.5, treat federated OAuth tokens as exposed to any script that ran in the LibreChat origin.
  4. Search logs from older versions for plaintext passwords from failed login validation and purge them.
03.

How to fix

  1. Upgrade LibreChat to 0.8.9 or later.
  2. Restrict who can add remote MCP servers, and block the LibreChat server from reaching internal/metadata addresses at the network level.
  3. If you ran 0.8.5, revoke and reissue federated OAuth/OpenID sessions.
docker ps --format '{{.Image}}' | grep -i librechat
# source install
grep '"version"' package.json
04.

Sources

Related on VibeVulns
The weekly email

One email a week. Only what changed.

Every new security hole in AI coding tools, MCP servers, app builders and the stacks they generate (Next.js, Supabase, Node, PHP), each with how to check your app and the fix. No roundups of things you already read.

Free, one email a week. Read the latest issue first if you like. One-click unsubscribe in every email. We never sell or share addresses.