Feast
feast.dev
- Free plan
- Linux
- Web
- API
- Self-hosted
What Feast is
Feast is an open-source feature store for delivering structured data to AI and LLM applications during training and inference. It supports machine learning feature management and serving for batch and real-time use. Point-in-time joins help keep future values out of model training data. Feature services support discovery, collaboration, and versioning of feature sets. Its Python SDK and CLI manage version-controlled feature definitions, materialize values, build training datasets, and retrieve online features. The Python feature server exposes features through an HTTP endpoint with JSON input and output, which clients in any language that can make HTTP requests can use. Feast describes integrations with offline and online stores and data sources, including community and custom integrations. It can be deployed on Kubernetes, where feature servers and jobs can run as workloads. Feast supports OIDC and Kubernetes RBAC authorization, but defaults to no_auth and leaves authentication token management to clients. Batch transformations require a separate transformation engine. Feast is free and open source.
Who it suits
Feast is designed for data scientists, MLOps engineers, data engineers, and AI engineers who need feature management and serving. It may suit teams working across batch and real-time applications.
Good to know, and what to watch
- Supports both batch and real-time feature serving
- Point-in-time joins help prevent training-data leakage
- Python SDK and CLI manage feature workflows
- Online and offline store support
- Free and open source
- Default authorization configuration is no_auth
- Clients must manage and pass authentication tokens
- Batch transformations require a separate engine
- Spark stream processor is experimental
Plans
Every plan and priceFeast
Open source
Getting started
- Open the Feast website
- Install or run Feast on a supported API, Linux, self-hosted, or web environment
- Use the Python SDK or CLI to define version-controlled features
- Configure offline and online stores or data sources
- Run the feature server and retrieve features through its HTTP JSON endpoint
- Deploy servers and jobs as Kubernetes workloads if required
The Club's review
Overview
Feast is an open-source feature store for teams serving structured data to AI and machine-learning applications. It is best suited to data, ML and MLOps practitioners coordinating features across training and inference. Its broad batch and real-time scope is useful, but teams need to supply authentication and a separate engine for batch transformations.
Key features
Point-in-time joins select feature values as of the relevant time, helping keep future information out of training data. That makes Feast a practical fit for teams building models from historical data, where leakage can undermine evaluation. Feature services support discovery and collaboration on existing features and version feature sets, helping teams manage shared definitions.
The Python SDK and CLI cover version-controlled feature definitions, materialization, training dataset creation and online feature retrieval. For serving, Feast's Python feature server exposes an HTTP endpoint with JSON input and output, so clients in other languages can request features without using the Python SDK.
Feast supports offline and online stores and multiple data sources, including community and custom integrations. It also supports feature monitoring. Kubernetes deployments can run feature servers and scheduled or ad-hoc jobs as workloads, which suits teams already operating that environment. Stream processing is a weaker point: the Spark processor is experimental and can consume from Kafka, so it is not the right choice for teams that need a mature, central stream-processing path.
Transformations are supported for on-demand and streaming sources, but batch transformations require a separate engine. Access control supports OIDC and Kubernetes RBAC, yet the default authorization configuration is no_auth. Feast does not provide authentication: clients must manage and pass tokens to the server. Teams should plan that responsibility rather than treat Feast as a complete, ready-made security layer.
Pricing
Feast is open source and free: its Feast plan costs nothing. The plan includes online and offline stores, point-in-time joins and feature monitoring, with both deployment models and both serving modes supported. That makes it a straightforward option for teams able to run and manage their own feature-store setup; the trade-off is that authentication and the separate batch transformation engine remain operational responsibilities.
Platforms
Feast supports API, Linux, self-hosted and web platforms. Its Python server's HTTP interface also allows feature requests from clients in any language that can make HTTP requests.
Who it's for
Feast is designed for data scientists, MLOps engineers, data engineers and AI engineers. It suits teams that need to share and version feature definitions while serving both training and inference workloads. Teams without capacity to manage client-side authentication or supply a separate batch transformation engine should look elsewhere.
Pros and cons
Pros
- Free and open source: the free plan includes both online and offline stores, feature monitoring and point-in-time joins.
- Consistent training and serving workflow: the SDK and CLI handle definitions, materialization, training datasets and online retrieval, while the HTTP server serves JSON requests across language boundaries.
- Deployment flexibility: both deployment models and serving modes are supported, with Kubernetes workloads available for servers and jobs.
Cons
- Authentication is external: clients must manage and pass tokens, and the default authorization setting is no_auth.
- Batch transformations need another engine: Feast does not cover that part of the workflow by itself.
- Experimental stream processing: the Spark processor's experimental status makes it a less dependable fit for teams centered on that capability.
Alternatives
Compare feature store software if you want to weigh Feast against a broader set of options.
- Canal is another free, open-source option for API, Linux, macOS and self-hosted use.
- Hopsworks Feature Store offers a free plan with one project, Feature Store and Model Registry, and community support; choose it if that included project is a fit. Its Enterprise plan uses custom pricing.
- OpenMLDB is a free open-source machine-learning database with standalone and cluster versions.
- JFrog ML Feature Store is paid and offers a free trial; choose it if you want to evaluate that option before buying.
- Databricks Feature Store offers usage-based monthly billing with per-second granularity and a free trial; choose it if that billing model suits your workload.
- Red Hat OpenShift AI Feature Store requires Red Hat OpenShift and offers self-managed subscriptions with Standard or Premium support; choose it if that deployment and support model fits.
- Chronon is a free Apache 2.0 open-source option with no stated usage limits.
Verdict
Choose Feast if your data or ML team wants a free, open-source feature store spanning point-in-time training data and online serving, and can own the surrounding infrastructure. Its main advantage is that breadth without a license fee; its main reason to look elsewhere is the need to handle authentication yourself and bring a separate engine for batch transformations.
The verdict
Feast brings feature management and serving for AI and machine learning applications into an open-source feature store. Teams should account for client-managed authentication and the separate engine needed for batch transformations.
Questions people ask
Is Feast free?
Yes. Feast is free and open source, with the Feast plan priced at no cost.
Which platforms does Feast support?
Listed platforms are API, Linux, self-hosted, and web.
Does Feast provide online and offline stores?
Yes. Feast supports both online and offline stores.
Does Feast handle authentication?
No. Clients are responsible for managing and passing authentication tokens to the server.
Can Feast serve features to other programming languages?
Yes. Its Python feature server exposes an HTTP endpoint with JSON input and output for any language able to make HTTP requests.
Does Feast support Kubernetes?
Yes. Feature servers and scheduled or ad-hoc jobs can run as Kubernetes workloads.
What the maker publishes
- Access control
- Feast supports OIDC and Kubernetes RBAC authorization, while its default authorization configuration is no_auth. source
- Authentication responsibility
- Feast does not provide authentication capabilities; clients are responsible for managing and passing authentication tokens to the server. source
- Batch and real-time
- Feast supports machine learning feature management and serving for both batch and real-time applications. source
- Community support
- The Feast homepage invites users to join its Slack community for support from Feast developers. source
- Deployment
- Feast can be deployed on Kubernetes, where feature servers and scheduled or ad-hoc jobs can run as Kubernetes workloads. source
- Feature server
- The Python feature server serves features through an HTTP endpoint with JSON input and output, usable from any language that can make HTTP requests. source
- Feature versioning
- Feast enables discovery and collaboration on existing features and versioning of feature sets through feature services. source
- Intended users
- The quickstart identifies data scientists, MLOps engineers, data engineers, and AI engineers as users Feast is designed to serve. source
- Point-in-time correctness
- Feast joins feature tables using point-in-time logic to prevent future feature values from leaking into model training data. source
- SDK and CLI
- The Python SDK and CLI manage version-controlled feature definitions, materialize values, build training datasets, and retrieve online features. source
- Stores and sources
- Feast docs describe integrations with offline and online stores and data sources, including community and custom integrations. source
- Stream processing
- Feast's component overview describes an experimental Spark processor that can consume data from Kafka. source
- Transformations
- The architecture docs say Feast supports transformations for on-demand and streaming sources, while batch transformations require a separate transformation engine. source
- What it does
- Feast is an open-source feature store that delivers structured data to AI and LLM applications for training and inference. source
Checked against the maker's pages on 30 Sept 2026.