Summary
TransformerOptimus SuperAGI up to version 0.0.14 exposes the get_vector_db_details endpoint in superagi/controllers/vector_dbs.py without proper authentication or access control.
A remote attacker can query this vector database management endpoint over the network and retrieve internal configuration details without valid credentials, increasing the risk of data exposure and follow-on attacks.
Affected context
The vulnerability occurs in deployments of TransformerOptimus SuperAGI (versions up to 0.0.14) where the Vector Database Management API is reachable over a network (e.g., from the internet, a corporate LAN, or a shared services environment).
The get_vector_db_details function in superagi/controllers/vector_dbs.py can be invoked without enforcing authentication or session checks.
This is especially risky in setups where SuperAGI is used to orchestrate AI agents against sensitive vector databases containing proprietary documents, embeddings, or connection metadata.
Exploit preconditions
- Attack surface: HTTP(S) Vector Database Management endpoint exposing
get_vector_db_details. - Required access: Network-level ability to reach the SuperAGI API (no valid user account or token required under the vulnerable conditions).
- Trigger point: Directly calling the
get_vector_db_detailsroute, typically via a crafted HTTP request, to retrieve vector database information. - Why exploitation is possible: The handler omits or misconfigures authentication/authorization checks (CWE-287/CWE-306), treating the request as trusted even when no identity or role has been established.
Why this happens
At its core, this issue is a missing or incomplete authentication and access-control check on a sensitive management endpoint.
The get_vector_db_details controller appears to:
- Accept incoming requests without verifying a valid user session, API token, or other identity proof, and/or
- Fail to enforce that only authorized roles can access vector database configuration data.
This maps to:
- CWE-287: Improper Authentication – the system does not adequately verify the identity of the requester, allowing unauthenticated access.
- CWE-306: Missing Authentication for Critical Function – a critical administrative/read-management function is exposed without authentication.
Conceptually, the bug class is a broken security boundary: an internal management capability (enumerating or describing vector databases) is exposed as if it were a public read-only resource.
Potential impact
- Unauthorized retrieval of vector database configuration details, including:
- Database types, names, or indexes
- Connection endpoints and internal service topology
- Exposure of sensitive metadata that may hint at the nature or structure of embedded documents.
- Increased success likelihood of targeted attacks (e.g., against backing databases or adjacent services) through reconnaissance.
- Potential chaining with other weaknesses (misconfigurations, weak credentials elsewhere) to gain deeper access to data or systems.
- Intelligence leakage that may reveal how AI agents are wired to internal data sources and workflows.
General mitigation guidance
Effective mitigation centers around enforcing strict authentication and authorization on all management endpoints and treating vector database configuration as sensitive information.
Common defensive patterns include:
- Centralized authentication middleware: Ensure every controller, including
vector_dbsroutes, is consistently wrapped with authentication middleware that validates user/session/API tokens. - Role- and scope-based access control: Require explicit privileges (e.g., admin/operator role or a specific "vector-db:read" scope) to access database management functions.
- Secure-by-default API design: Treat all non-trivial endpoints (especially those disclosing configuration details) as private by default, requiring opt-in exposure and explicit security annotations.
- Network and deployment hardening: Restrict SuperAGI's management interfaces to trusted networks or service meshes; avoid exposing them directly to the public internet.
- Least-privilege information disclosure: Reduce the granularity of data returned by such endpoints to the minimum necessary for legitimate use, avoiding secrets or unnecessary infrastructure details.
How teams detect issues like this
Organizations can detect and prevent similar issues using a combination of code-level and operational strategies:
- Static analysis and linters:
- Rules that flag controller functions or routes that lack required authentication decorators/middleware.
- Patterns that detect public routes mapped to "admin", "manage", "config", or "db" semantics without explicit access control checks.
- Secure code review practices:
- Review API controllers for consistent use of authentication/authorization, especially around configuration, database, or admin endpoints.
- Check that all routes under management namespaces (e.g.,
/admin,/vector_dbs) enforce role-based access. - API gateway and policy enforcement:
- Use API gateways or service meshes to enforce authentication policies (e.g., JWT validation, mTLS) regardless of application code.
- Apply route-based policies that block unauthenticated access to sensitive paths.
- Runtime monitoring:
- Monitor for anomalous access patterns to management endpoints from unknown or untrusted IP ranges.
- Use WAF rules to detect enumeration-style behavior (e.g., repeated unauthenticated calls to vector database endpoints).

