Skip to the content
Know your tech before you buy it. Makers: get listed from $49/mo

Agent Platform Feature Store

cloud.google.com

  • Web
  • API
6.0 Club score Visit cloud.google.com

What Agent Platform Feature Store is

Agent Platform Feature Store is a paid, cloud-native managed service for storing machine-learning features and serving current values online from BigQuery sources. BigQuery tables and views provide the offline store, so the service does not create a separate offline repository. For online predictions, it serves latest feature values at low latency. Teams may optionally register sources and columns in the Feature Registry to select fields, combine sources, and monitor statistics or drift. Monitoring runs manually or on a schedule and can detect anomalies such as feature drift. Bigtable online serving handles large volumes and near-real-time continuous BigQuery synchronization but does not support embeddings. Optimized online serving supports scheduled synchronization, is deprecated, and is scheduled to sunset on February 17, 2027. IAM controls access to feature groups, online store instances, and views. Resources must share a region or multi-region with BigQuery sources. There is no online-store retention limit; BigQuery limits may apply offline.

Who it suits

It suits machine-learning teams serving BigQuery-based features for online predictions. It is relevant to organizations needing optional registry, monitoring, IAM, and cloud deployment controls.

Good to know, and what to watch

  • Managed online serving from BigQuery
  • BigQuery tables and views serve as offline store
  • Feature monitoring detects statistics and drift
  • IAM policies control feature resource access
  • Supports point-in-time joins and both serving modes
  • No free plan is listed
  • Optimized online serving is deprecated
  • Optimized serving lacks continuous synchronization
  • Bigtable serving does not support embeddings

New Agent Platform Feature Store (pay-as-you-go)

—

The verdict

Agent Platform Feature Store links BigQuery data with managed feature serving, monitoring, and access control. Teams should account for the deprecated optimized mode, regional placement rules, and separate Vector Search guidance for embeddings.

What the maker publishes

Data retention
The online store has no data retention limit, while BigQuery retention limits or quotas may apply to the offline data. source
Data sync limit
Optimized online serving supports scheduled data sync but not continuous data sync; Bigtable online serving supports near-real-time continuous sync from BigQuery. source
Embeddings
Google recommends its purpose-built Vector Search to efficiently store and serve embeddings. source
Feature registry
Registering data sources and feature columns in the Feature Registry is optional; registration supports selecting columns, combining sources, and monitoring statistics and drift. source
IAM access control
IAM policies can control access to feature groups, online store instances, and feature views, with bindings for users, groups, domains, and service accounts. source
Location constraint
Feature Store resources must be in the same region or multi-region as the BigQuery source, and dual-region bucket sources are unsupported. source
Monitoring
Feature monitoring can run on a schedule or manually to retrieve statistics and detect anomalies such as feature drift. source
Offline data
BigQuery tables and views form the offline store, so Feature Store does not provision a separate offline store. source
Online predictions
It serves the latest feature values from BigQuery for online predictions at low latencies. source
Purpose
It is a managed, cloud-native feature store for managing ML features and serving them online from BigQuery data sources. source
Serving options
Bigtable online serving supports large data volumes but not embeddings; optimized online serving is deprecated and scheduled to sunset on February 17, 2027. source
Support and learning
The documentation links to Google Cloud Support, community forums, release notes, and tutorials with Colab and GitHub examples. source

Checked against the maker's pages on 4 Oct 2026.