CVE-2026-71424: Onyx: Cross-user OAuth-token leak via /api/mcp/servers* for per-user MCP servers
Onyx is an open-source AI platform. Prior to 3.1.10, 3.2.14, and 4.0.0, Onyx's GET /api/mcp/servers and GET /api/mcp/servers/persona/{persona_id} endpoints expose another user's OAuth Authorization header because OnyxTokenStorage.set_tokens and OnyxTokenStorage.set_client_info in backend/onyx/server/features/mcp/api.py copy per-user tokens into a shared admin MCPConnectionConfig row and _db_mcp_server_to_api_mcp_server returns that row through auth_template.headers to any BASIC_ACCESS user. This issue is fixed in versions 3.1.10, 3.2.14, and 4.0.0.
Security readout for executives and security teams
Plain-English summary
A signed-in, basic-access Onyx user could receive another user’s OAuth authorization token through affected MCP server API responses. A stolen token could expose connected services or permit actions within that token’s granted permissions. Onyx has released corrected versions.
Executive priority
Treat this as an urgent upgrade and credential-containment issue. Prioritize internet-reachable or multi-user Onyx deployments with OAuth-enabled MCP integrations. Potential impact crosses user boundaries and may extend into connected services, although actual compromise is not established by the supplied evidence.
Technical view
Per-user OAuth tokens were copied into a shared administrative MCPConnectionConfig row. Two GET endpoints then returned that shared row through auth_template.headers to BASIC_ACCESS users, breaking user-level authorization boundaries. The affected endpoints are /api/mcp/servers and /api/mcp/servers/persona/{persona_id}.
Likely exposure
Exposure applies to Onyx versions below 3.1.10; versions 3.2.0 through 3.2.13; and versions 3.3.0-beta.0 through 4.0.0-exclusive. Risk is highest where per-user MCP servers use OAuth and untrusted or broadly provisioned BASIC_ACCESS accounts can reach the affected APIs.
Exploitation context
The provided sources do not report active exploitation, and the source bundle marks KEV as false. Exploitation nevertheless requires only network access and a low-privileged authenticated Onyx account, with no user interaction, according to the CVSS vector.
Researcher notes
The root issue combines sensitive-data exposure with missing user-level authorization enforcement. CVSS 3.1 is 9.6: network-accessible, low complexity, low privileges, no interaction, changed scope, and high confidentiality and integrity impact. Availability impact is rated none. Assess token scope and endpoint-access evidence when estimating realized impact.
Mitigation direction
Upgrade affected 3.1 deployments to version 3.1.10 or later.
Upgrade affected 3.2 deployments to version 3.2.14 or later.
Upgrade affected beta or newer branches to version 4.0.0 or later.
Revoke and rotate potentially exposed OAuth tokens after assessing the affected period.
Restrict BASIC_ACCESS accounts and affected endpoint access until upgrading is complete.
Validation and detection
Inventory deployed Onyx versions and compare them with the documented affected ranges.
Identify per-user MCP servers configured with OAuth during the affected period.
Review available application or proxy logs for access to the affected endpoints.
Confirm upgraded API responses no longer disclose OAuth Authorization headers across users.
Verify potentially exposed tokens were revoked and replacement credentials are functioning.
Generated from the cited source records. This long-tail analysis has not been individually reviewed by a named human.
Potential ATT&CK relevance
Conservative CVE-to-ATT&CK context
These mappings and lookup hints may be relevant to the vulnerability behavior, CWE, affected product, or exposure path. Glexia-inferred context is not an official MITRE, ATT&CK, CWE, or CVE Program mapping.
ATT&CK lookup starting points
Use these exact CWE pages and searches to review the Glexia ATT&CK library from this CVE's weakness and description context.
cwe · medium confidence lookup
CWE-200: Information exposure and cloud metadata lookup
Information exposure and SSRF weaknesses can make discovery, cloud metadata, and credential material review relevant. Open the exact CWE lookup page first, then review the ATT&CK searches from that MITRE weakness context. This is a Glexia lookup hint, not an official ATT&CK mapping.
CWE-863: Authorization and privilege behavior lookup
Authorization weaknesses can support privilege escalation and valid-account review, depending on exploit path. Open the exact CWE lookup page first, then review the ATT&CK searches from that MITRE weakness context. This is a Glexia lookup hint, not an official ATT&CK mapping.
The CVE wording references authentication or credential exposure, so valid-account and credential-access review may help. This is a Glexia inferred lookup path, not an official MITRE, ATT&CK, or CVE Program mapping.
These fields come from the CVE record and ADP containers, not from Glexia's Take. They preserve time-varying source decisions such as CISA SSVC, KEV status, CVSS metrics, and provider references.
1CVSS vectors
3Timeline events
1ADP providers
9Source links
SSVC decision data
CISA-ADPCISA Coordinator
Timestamp
Version
2.0.3
Exploitation: noneAutomatable: noTechnical Impact: total
CVSS vector scores
1 official score
We collect every scored CVSS vector available in the official CNA and ADP containers. When more than one version is present, the table keeps the source vectors side by side instead of collapsing them into the highest score.
CWE links open Glexia weakness intelligence pages with official CWE context, developer remediation guidance, and related CVE mappings.
CWE-200 · source CWE mapping
Exposure of Sensitive Information to an Unauthorized Actor
Exposure of Sensitive Information to an Unauthorized Actor represents a recurring weakness pattern that can create exploitable paths when design, validation, or implementation controls are missing.
Incorrect Authorization represents a recurring weakness pattern that can create exploitable paths when design, validation, or implementation controls are missing.