Integration · Feature store
Feast + OrchestrAI
Catalog exported 2026-09-02 · Feast website
Apply Feast definitions, materialize features, and fetch online or historical features from chat.
OrchestrAI exposes 10 Feast operations: 7 are low-risk (read-only or low-impact), and 3 create or modify resources and run only after you confirm the plan.
What teams use it for
ML engineers use OrchestrAI with Feast to check which feature views exist, materialize the latest values into the online store before a model release, and spot-check online features for a specific entity key. Applying definitions and both materialize operations are medium risk and run when requested, while every list and get is read-only. There is no operation to delete feature views or entities, and no push source ingestion, so cleanup and streaming writes still happen with the Feast CLI or SDK.
Every Feast operation, with its risk level
| Operation | What it does | Risk | Step-level approval |
|---|---|---|---|
Get Feast Feature Service |
Get feature service details | Low risk | No |
Get Feast Feature Store Info |
Get feature store info | Low risk | No |
Get Feast Historical Features |
Get historical features for training | Low risk | No |
Get Feast Online Features |
Get online features for real-time inference | Low risk | No |
List Feast Data Sources |
List data sources | Low risk | No |
List Feast Entities |
List entities | Low risk | No |
List Feast Feature Views |
List feature views | Low risk | No |
Apply Feast Feature Store |
Apply feature store definitions | Creates resources | No |
Materialize Feast Features |
Materialize features to online store | Creates resources | No |
Materialize Feast Features Incrementally |
Incrementally materialize features | Creates resources | No |
Risk tiers come from the catalog: low is read-only or low-impact, medium creates resources and is reversible, high modifies existing resources, destructive may lose data. Every plan that creates or changes resources is shown with its cost estimate and waits for your confirmation. Operations marked with a step-level approval pause again on their own step. Destructive operations require a typed risk phrase.
What you connect
A Feast credential (stored as feast).
Connected-service tokens are envelope-encrypted with a per-record key wrapped by a cloud KMS.
Prompts that work
- List the feature views in the Feast repo and show which ones feed the fraud_v3 feature service
- Materialize features incrementally up to now for the online store
- Get online features for driver_id 1042 from the driver_hourly_stats view
Before anything runs
Every mutation shows its plan, cost estimate, and blast radius, then waits for your confirmation. Destructive operations require a typed risk phrase. Credentials are minted per run through OIDC federation and discarded afterward; nothing you create here is invisible later, because every resource lands in the desired-state ledger where drift is detected and can be converged. Details on the security page.
Frequently asked questions
- Can OrchestrAI materialize Feast features to the online store?
- Yes. Full and incremental materialization are medium-risk operations that run on request and write to the configured online store.
- Does OrchestrAI remove Feast feature views?
- No. There is no delete operation in the Feast integration. Applying definitions can add or update objects, but removal is done outside OrchestrAI.
- How does OrchestrAI authenticate to Feast?
- You add a Feast credential once in the connections screen. It is envelope-encrypted with a per-record key wrapped by a cloud KMS and is only decrypted inside the run that needs it.
Related integrations
Try it on your own account
Connect your cloud read-only and see your resources, drift, and costs before anything runs. $5 minimum to start. Unused credits refunded in your first 14 days.
Unused credits refunded in your first 14 days.