B — Some risk signals Graded October 16, 2026Minor concerns worth a quick check.Based on public adoption and transparency signals as of October 16, 2026. Our opinion, not a code audit or security guarantee.
App connector · Data & databases
whodb
Where data access meets operational intelligence
Bottom line: Signal-based grade (4/7 points): very widely starred on GitHub, public source code. This reflects adoption and transparency, not a code review.
- Popularity score
- 42/100
- Rank
- #280 of 500
- GitHub stars
- 5k
What it does
WhoDB is an MCP connector that provides SQL database access and operational intelligence capabilities, allowing AI agents to query and interact with databases through a standardized interface. It bridges database systems with AI workflows, enabling data-driven decision-making and real-time operational visibility.
Who it’s for
Data engineers, platform builders, and executives who need AI agents to access structured data for analytics, reporting, or automated operations. Teams running Claude or other MCP-compatible AI systems that require secure database query capabilities.
Copy-paste setup
Check these steps against the project’s own instructions before running anything.
Verify against the repo README at https://github.com/clidey/whodb (opens external site) for exact MCP server setup instructions. Likely involves installing via npm or Docker, then configuring database credentials and connection strings in your Claude Desktop or MCP host configuration. Exact snippet requires checking repo documentation.
Risks
- Credential scope: Database credentials passed to AI agents create broad data access risk; SQL injection or prompt injection could expose sensitive schemas or data
- Data exfiltration: Agents may return full result sets to external APIs or logs if not properly sandboxed
- SQL injection: Malicious prompts could be converted into destructive queries (DROP, DELETE) if input sanitization is insufficient
- Audit trail gaps: Database access via AI intermediary may bypass traditional access logging and compliance controls
- Supply chain: Low npm adoption signals suggests limited production vetting; maintenance status should be verified before enterprise deployment
How to stay secure
How to Stay Secure with whodb
Least-privilege permissions
- Create a dedicated database user for the AI agent with access only to the specific tables it needs—not the entire database or sensitive schemas like user passwords or financial records.
- Use read-only access whenever possible; prevent the agent from executing commands that could modify or delete data (DROP, DELETE, UPDATE statements).
- Rotate credentials regularly (every 30–90 days) and use unique passwords that are randomly generated, not reused across systems.
- Avoid storing credentials in code or config files; use a secrets vault (encrypted storage system) that only the agent can access.
Network isolation
- Run the agent in a restricted network zone separate from production databases; use a firewall to allow connections only between the agent and its authorized database.
- Disable public internet access from the agent to external APIs unless explicitly required; this prevents stolen credentials or data from being sent to unknown servers.
- Use encrypted connections (SSL/TLS) for all communication between the agent and the database to prevent credential or data interception over the network.
Monitoring
- Log every query the agent runs, including what data was requested and by whom; compare logs weekly against expected activity patterns.
- Set alerts for unusual patterns—multiple failed login attempts, requests for sensitive tables, large data downloads, or commands happening outside business hours.
- Audit the MCP connector regularly for security patches and maintenance updates; verify the software is actively supported by its developers before use.
Red flags to watch for
- Unexplained large data exports or queries returning thousands of rows when the agent typically returns small result sets.
- Access to tables the agent shouldn't need, particularly authentication, payment, or personal information tables.
- Missing or incomplete audit logs for agent activity—gaps in logging make it impossible to detect or investigate breaches.
- Software no longer maintained; check the GitHub repository or npm package page for recent updates; if the last commit was >6 months ago, it may contain unpatched security flaws.
Alternatives
- Anthropic's native SQL tool – First-party database integration with tighter security controls
- Temporal.io MCP connector – Workflow-focused alternative for operational data without direct database exposure
- DuckDB + Parquet connectors – Lighter-weight, file-based approach for analytics without live database risk
- Retool / Internal tools platforms – Pre-built UI and access controls reduce AI-mediated database risk