DEV Community

developerz.ai
developerz.ai

Posted on

Securing AI Agent Database Access with db-mcp-gateway

Securing AI Agent Database Access with db-mcp-gateway

Published on July 31, 2026

Introduction

As AI agents become integral to modern applications, they often need direct access to production databases for real‑time analytics, feature extraction, or autonomous decision‑making. However, exposing raw database credentials to these agents introduces a massive attack surface. db-mcp-gateway solves this problem by providing a self‑hosted Model Context Protocol (MCP) gateway that isolates credentials, enforces SSO‑based authentication, and records a complete audit trail for every query.

Core Security Principles

Credential Isolation

  • Never expose database URLs – credentials live only inside the gateway container.
  • AI agents receive only query results, never connection strings.
  • No credential leakage appears in logs, errors, or responses. 
AI Agent → MCP Protocol → Gateway → Database
          ↓                ↓           ↓
   No Credentials      Auth Only    Least Privilege
Enter fullscreen mode Exit fullscreen mode

Identity & Access Control

The gateway integrates with popular identity providers (Okta, Google Workspace, Entra, Authentik, Keycloak) using a browser‑based SSO flow—no embedded browsers required. Permissions are defined as grant‑based YAML files that can be reviewed via pull requests, enabling GitOps‑style management.

grants:
  - group: backend-devs
    databases: [production_postgres]
    actions: [query_read]
    constraints:
      schemas: [public, analytics]
      row_limit: 1000
      require_reason: true
Enter fullscreen mode Exit fullscreen mode

Audit Trail

Every query is recorded with the originating SSO user, group, grant, and a timestamp. This immutable log satisfies compliance auditors needing evidence of who accessed what data and when.

Key Features

Feature Description
MCP Tool Surface list_databases, describe_schema, sample_table, run_query, explain, get_query_history
SSO Integration Supports Okta, Google Workspace, Entra, Authentik, Keycloak
Security Model Read‑only by default; writes opt‑in per grant; per‑database least‑privilege roles; statement timeouts and row caps
Deployment Single Docker container, one YAML config, PostgreSQL for state and audit logs; supports PostgreSQL & MongoDB

Example Deployment 

# Pull the latest image
docker pull ghcr.io/developerz-ai/db-mcp-gateway:1.1.1

# Run with your config
docker run -p 8080:8080 \
  -v $(pwd)/config.yaml:/app/config.yaml \
  ghcr.io/developerz-ai/db-mcp-gateway:1.1.1
Enter fullscreen mode Exit fullscreen mode

The container starts a secure gateway listening on port 8080. All AI agents must authenticate via the configured SSO provider before issuing any MCP commands.

Why Choose db-mcp-gateway?

  • Security First – Credentials never leave the gateway, eliminating accidental leaks.
  • Compliance Ready – Full audit trails and SSO integration simplify meeting internal security policies.
  • Fine‑Grained Control – Grants can limit schemas, rows, and even require a justification for each query.
  • GitOps‑Friendly – Permissions live as code, enabling peer review and version control.

Getting Started

  1. Clone the repository: git clone https://github.com/developerz-ai/db-mcp-gateway.git
  2. Adjust config.yaml with your SSO provider and desired grants.
  3. Deploy the Docker container and point your AI agents to http://<gateway-host>:8080.
  4. Use the run_query command to safely execute SELECT statements.

Conclusion

`-mcp-gateway provides a pragmatic, security‑first approach to granting AI agents database access without compromising credential secrecy. By leveraging SSO, grant‑based permissions, and immutable audit logs, it empowers Platform/SRE teams, backend developers, and security officers to adopt AI‑driven workflows with confidence.

Explore the project and start a secure deployment today: https://github.com/developerz-ai/db-mcp-gateway

Keywords: #MCP #DatabaseSecurity #AI #Compliance #PostgreSQL

Top comments (0)