Database Connectors
Connect MUJPI securely to your organization's databases. While direct database access by AI models carries inherent risk, MUJPI's database connectors wrap queries in strict, policy-enforced layers, prioritizing read-only analytics and tightly controlled write operations.
Architecture & Data Flow
AI requests are translated into queries, passed through an authorization layer to ensure compliance with field-level restrictions, and executed against the database.
Data Flow
Queries are executed via native database protocols (e.g., Postgres wire protocol). Results are paginated, sanitized to remove sensitive fields, and returned to the agent context.
Typical Use Cases
- ✦Natural language querying of read-only analytical databases
- ✦Retrieving customer context during support workflows
- ✦Generating operational reports from live metrics
- ✦Controlled execution of pre-approved stored procedures
Capabilities
- ✦PostgreSQL, MySQL-compatible, and Microsoft SQL Server support
- ✦Schema introspection and mapping
- ✦Query execution with strict timeouts and limits
- ✦Connection pooling
Security Considerations
- Default read-only connectivity is strongly enforced.
- Schema mapping strictly limits which tables and columns the AI can perceive.
- Query execution timeouts and row-return limits prevent denial of service.
- Write access is restricted to specific, pre-approved parameterized procedures.
Setup & Deployment
Deployment Modes
Setup Overview
Configure connection strings with least-privilege credentials. Define schema visibility rules and query limits in the MUJPI administration console.