Google Cloud and Oracle have released three tightly coupled capabilities: a Remote Model Context Protocol (MCP) Server backed by the ExecuteSQL Data API, a preview integration with Google Cloud Knowledge Catalog, and a Gemini connector that lets users query Oracle AI Database@Google Cloud in natural language. Together they replace ad‑hoc scripting, bastion hosts, and custom credential handling with a unified, IAM‑controlled interface for agentic workloads.
Remote MCP Server and ExecuteSQL Data API
The Remote MCP Server is now generally available. It exposes a Data API that executes SQL against Oracle databases while automatically resolving credentials through Secret Manager and enforcing Cloud IAM policies. For engineers, this means:
- Elimination of custom "natural‑language‑to‑SQL" middleware; agents can call the API directly.
- Reduced surface area for credential leakage because passwords are never stored in code or environment variables.
- Consistent audit trails via IAM logs, simplifying compliance reporting.
Implementation wise, services that need database access must be granted the appropriate IAM role on the Data API and be configured to read the secret reference from Secret Manager. The API is a standard HTTPS endpoint, so existing service meshes or egress controls can be reused.
Knowledge Catalog preview integration
Google Cloud’s Knowledge Catalog now offers a preview that automatically discovers Oracle schemas and business metadata. The catalog creates a unified view of metadata across Oracle and other Google data services, enabling:
- Centralized governance policies that can be applied to both Oracle and BigQuery assets.
- Provenance tracking for data accessed by AI agents, which helps trace the origin of model inputs.
- Rich semantic context that agents can use to disambiguate column meanings during query generation.
Practitioners should treat the preview as an optional add‑on that requires sign‑up. Once enabled, the catalog populates entries without manual tagging, but the underlying access controls still rely on the IAM policies governing the Remote MCP Server.
Gemini connector preview
The Gemini connector adds a native integration inside the Gemini application, allowing users to type natural‑language questions that are resolved against Oracle AI Database@Google Cloud. Behind the scenes Gemini leverages Knowledge Catalog to discover the relevant Oracle objects and enforces the same IAM‑based data permissions as the Remote MCP Server. This provides a conversational front‑end without exposing raw SQL or bypassing security controls.
From an operational perspective, the connector is a preview feature; it should be evaluated in a non‑production environment before wider rollout. Monitoring should focus on API latency, IAM audit logs, and any unexpected data access patterns.
Related CloudNinjas coverage: Google Cloud.
What This Means For Practitioners
Engineers should start by mapping existing Oracle data pipelines to the Remote MCP Server and ExecuteSQL Data API, replacing any custom credential handling with Secret Manager references. Enable the Knowledge Catalog preview where metadata governance is a priority, and test the Gemini connector in a sandbox to validate conversational query accuracy and permission enforcement. Ongoing considerations include tracking IAM role assignments, reviewing audit logs for anomalous access, and planning for the eventual GA of Knowledge Catalog to avoid re‑architecting later.
