Servers and applications
CVE-2026-54044
Shared-Agent Editor Can Exfiltrate Owner Action API Keys via OpenAPI Spec Swap
- Severity
- High 7.7
- Published
- 17 Nov 2026
- Last changed
- 17 Nov 2026
- Added to UpHeron
- 17 Nov 2026
- Fixed in
- 0.8.7
- Attack
- Over the network · Needs a user account · No user action needed
- Attacker can
- Read data
- CVSS vector
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:C/C:H/I:N/A:N- Also known as
- GHSA-qh43-24gm-5j57
- Sources
- Project advisories (GitHub)
Description
LibreChat's Actions feature lets an agent owner attach an OpenAPI-described tool and store an API key for it (metadata.auth.api_key, encrypted at rest). LibreChat sends that key as Authorization: Bearer whenever the LLM calls the tool. Owners can share an agent with collaborators, including at EDIT permission.
A collaborator with EDIT permission can send a single POST /api/agents/actions/:agent_id that swaps the action's OpenAPI spec and domain to a host they control, while the owner's encrypted api_key (and any oauth_client_id / oauth_client_secret) stays attached to the same action record. The next time the agent runs the tool, LibreChat decrypts the owner's credential and sends it to the attacker's host.
The editor never has to read or invoke anything. Any later invocation by anyone with VIEW or above — including the owner — leaks the secret. No exotic config is needed: agent sharing with EDIT and bearer-key Actions are both documented, default-path features.
Affected software
| Software | Affected versions |
|---|---|
LibreChatServers and applications · github:librechat-ai/librechat |
|
About LibreChat
Enhanced ChatGPT Clone: Features Agents, MCP, Skills, DeepSeek, Anthropic, AWS, OpenAI, Responses API, Azure, Groq, o1, GPT-5, Mistral, OpenRouter, Vertex AI, Gemini, Artifacts, AI model switching, message search, Code Interpreter, langchain, DALL-E-3, OpenAPI Actions, Functions, Secure Multi-User Auth, Presets, open-source for self-hosting. Active
- Docker Hub pulls
- 669,094Docker Hub, 15 Nov 2026
- GitHub stars
- 47,138GitHub, 15 Nov 2026
- Forks
- 9,706GitHub, 15 Nov 2026
- License
- MITGitHub
- First published on GitHub
- 12 Feb 2023GitHub
Source code Website Docker image
Vulnerabilities in LibreChat
- In our list
- 104
- Published in the last 12 months
- 84
- Critical, last 12 months
- 5
- High, last 12 months
- 25
- Known exploited
- 0
- CVE-2026-54041Plaintext password logged on login validation errorLow 3.1 17 Nov 2026
- CVE-2026-54032Remote MCP Server Description XSS via Unsanitized dangerouslySetInnerHTML; secondary Admin Banner sink lacks DOMPurifyMedium 6.1 17 Nov 2026
- CVE-2026-54026Cross-User Message Injection via Parameter Confusion in LibreChatMedium 6.5 17 Nov 2026
- CVE-2026-54031MCP OAuth requests can follow redirects to internal URLs after SSRF validationMedium 6.3 17 Nov 2026
- CVE-2026-54038Server-Side ReDoS via Unescaped Regex in Principal SearchMedium 6.5 17 Nov 2026
- CVE-2026-54045Password Reset Can Wipe Email-Verification Tokens When Verification Is EnforcedMedium 6.5 17 Nov 2026
References
- CVE.org https://www.cve.org/CVERecord?id=CVE-2026-54044
- GitHub Advisory Database https://github.com/advisories/GHSA-qh43-24gm-5j57
- github.com https://github.com/LibreChat-AI/LibreChat/security/advisories/GHSA-qh43-24gm-5j57