AlloyDB AI

What it does

Agent connectivity

AlloyDB includes a fully managed remote MCP server for providing agents access to data.checked 9 Oct 2026

AI functions

AlloyDB AI functions let SQL queries generate embeddings, perform similarity searches, call AI models, analyze sentiment, generate content, and run forecasts.checked 9 Oct 2026

Certifications

Google lists AlloyDB certifications including ISO 27001, ISO 27017, ISO 27018, ISO 27701, SOC 1, SOC 2, SOC 3, PCI DSS, and HIPAA.checked 9 Oct 2026

Deployment

AlloyDB Omni is a downloadable AlloyDB edition for on-premises, air-gapped, edge, or multicloud deployments.checked 9 Oct 2026

Encryption

AlloyDB encrypts cluster data at rest by default and supports customer-managed encryption keys, Cloud KMS Autokey, and Cloud External Key Manager.checked 9 Oct 2026

Framework integrations

The product page names Agent Development Kit, LangChain, and LlamaIndex as orchestration framework integrations.checked 9 Oct 2026

Model access

Users can access Gemini or other foundation models hosted on Gemini Enterprise Agent Platform and register endpoints on any platform for calls from AlloyDB SQL functions.checked 9 Oct 2026

Natural language

QueryData lets data agents translate natural-language questions into SQL and disambiguate questions when needed.checked 9 Oct 2026

Network security

AlloyDB supports TLS 1.3, VPC Service Controls, and private connectivity options; its default endpoint is not publicly routable.checked 9 Oct 2026

Notable pricing limit

CPU and memory are priced per vCPU and GiB, pricing varies by location, and an instance can have up to 288 vCPUs and 2,232 GiB memory per node.checked 9 Oct 2026

Purpose

AlloyDB AI integrates vector embeddings, high-performance vector search, and natural language into a PostgreSQL-compatible database for building AI agents and apps.checked 9 Oct 2026

SQL and vector queries

The ScaNN index integrates with the PostgreSQL query planner, and adaptive filtering coordinates filters, joins, and vector indexes.checked 9 Oct 2026

Vector search

Its ScaNN index supports pgvector-compatible semantic search and is described as scaling to over 10 billion vectors.checked 9 Oct 2026

Who it is for

The product is presented for developers building AI applications and agents grounded in structured and unstructured enterprise data.checked 9 Oct 2026