Agent Platform Feature Store review
A Google Cloud feature store with strong serving and monitoring, but a narrow ecosystem.
Reviewed by iTechGuides Editors · Editorial team · Updated Oct 2026
Agent Platform Feature Store, documented as Vertex AI Feature Store, is a managed cloud service for organizing, storing, monitoring, and serving machine-learning features. It is aimed at mid-market and enterprise teams working in Google Cloud, with support for both offline training data and online inference. BigQuery tables or views provide the offline store, while online values can be served through Cloud Bigtable or optimized online-serving infrastructure. Feature groups, a feature registry, and feature views help organize registered features and BigQuery sources.
Pricing is structured around processing, serving, and storage resources. The Data processing node covers ingestion into online stores and monitoring. The Optimized online serving node supports low-latency serving and embeddings serving, with 200GB of storage per node. The Bigtable online serving node provides serving with Cloud Bigtable, while Bigtable online serving storage covers storage for that serving option. The published starting price is 0.08, and offline operations also incur applicable BigQuery charges. There is no free plan.
Feature workflows extend from training data preparation to production retrieval. Historical point-in-time feature retrieval supports training and batch prediction, while current-value serving supports online inference. Scheduled and manual monitoring jobs provide feature statistics and drift detection, and optimized serving adds embedding management and vector retrieval. The main fit is a Google Cloud team that wants managed feature storage and serving within a BigQuery and Bigtable-centered environment. Teams seeking a cloud-agnostic deployment model, a broader integration ecosystem, or a long-term dependency on optimized online serving should consider alternatives, especially because Google Cloud states that this serving option will be fully sunset on February 17, 2027.
Agent Platform Feature Store pros and cons
- Where it wins
- BigQuery-backed offline storage with historical point-in-time retrieval
- Online serving through optimized infrastructure or Cloud Bigtable
- Feature monitoring includes statistics, scheduled jobs, and drift detection
- Where it doesn't
- Cloud-only deployment limits platform flexibility
- Integration set centers on BigQuery, Cloud Bigtable, and Knowledge Catalog
- Optimized online serving is scheduled for sunset on February 17, 2027
Agent Platform Feature Store fact sheet, pricing and score →
Advertiser disclosure: iTechGuides is reader-supported. Vendors can pay for top positions in our rankings and for a place on other products' pages, and we may earn a commission when you click some links. Bought places stand at the number bought; this page and our disclosure say so. Paying vendors may get a fuller, more detailed review from our editors, and that review can change their score. How we rank.
Last updated · How we research and update