Hopsworks Feature Store

Open-source ML platform combining a feature store, model registry, and model serving. Hopsworks provides batch and real-time feature pipelines, serverless feature computation, model versioning, and deployment. REST API and Python (HSFS) SDK for feature retrieval and model management. Hopsworks Serverless (managed cloud) provides a free tier for prototyping.

Evaluated Mar 06, 2026 (0d ago) vv3.x
Homepage ↗ Repo ↗ AI & Machine Learning feature-store mlops model-registry open-source python spark flink real-time serverless
⚙ Agent Friendliness
59
/ 100
Can an agent use this?
🔒 Security
83
/ 100
Is it safe for agents?
⚡ Reliability
78
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
--
Documentation
80
Error Messages
74
Auth Simplicity
85
Rate Limits
75

🔒 Security

TLS Enforcement
100
Auth Strength
80
Scope Granularity
75
Dep. Hygiene
82
Secret Handling
80

SOC2. HTTPS enforced. Project-scoped API keys. AGPL license requires open-source compliance review. EU data residency by default on Serverless. Self-hosted: security posture is operator-managed.

⚡ Reliability

Uptime/SLA
82
Version Stability
78
Breaking Changes
75
Error Recovery
78
AF Security Reliability

Best When

You want an all-in-one open-source ML platform with feature store, model registry, and serving — especially if you want a managed free tier for prototyping.

Avoid When

You're deeply embedded in AWS/Azure/GCP and prefer native managed services — vendor-native MLOps platforms will integrate more smoothly.

Use Cases

  • Build end-to-end ML pipelines with feature engineering, training, and serving using Hopsworks as a unified platform
  • Serve real-time features to agent inference with online feature store backed by RonDB (MySQL NDB Cluster)
  • Version and deploy ML models via Hopsworks model registry with automatic deployment integration
  • Define feature groups with validation and drift detection for agent model quality assurance
  • Use Hopsworks Serverless free tier to prototype feature store patterns before committing to self-hosted infrastructure

Not For

  • Enterprise teams needing FedRAMP compliance or specific certifications — verify current compliance status
  • Teams already on Databricks or AWS — vendor-native feature stores (Databricks Feature Store, SageMaker) integrate better in those ecosystems
  • Simple batch ML without real-time serving — lighter alternatives (Feast) are simpler to operate

Interface

REST API
Yes
GraphQL
No
gRPC
No
MCP Server
No
SDK
Yes
Webhooks
No

Authentication

Methods: api_key bearer_token
OAuth: No Scopes: No

API keys generated in Hopsworks UI. Keys used for SDK authentication and REST API access. Project-scoped keys — each project has its own key set. Hopsworks Serverless uses the same key mechanism.

Pricing

Model: open_source
Free tier: Yes
Requires CC: No

Open-source core is free (AGPL). Hopsworks Serverless offers a generous free tier for prototyping. Enterprise license for commercial on-premise with support.

Agent Metadata

Pagination
page
Idempotent
Full
Retry Guidance
Not documented

Known Gotchas

  • AGPL license requires careful review for SaaS products — commercial use of self-hosted Hopsworks may require enterprise license
  • Feature group versions are immutable — schema changes require new feature group version with migration
  • Online store relies on RonDB — not available in all deployment configurations; verify online serving setup
  • Feature pipeline (data source → feature group) requires separate compute (Spark/Flink/Python) — not managed by Hopsworks in open-source
  • SDK must match server version — significant version drift causes API incompatibility
  • Hopsworks Serverless has EU data residency by default — US residency requires specific configuration
  • Model serving in Hopsworks uses KFServing internally — model container format must be compatible

Alternatives

Full Evaluation Report

Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for Hopsworks Feature Store.

$99

Scores are editorial opinions as of 2026-03-06.

5178
Packages Evaluated
26151
Need Evaluation
173
Need Re-evaluation
Community Powered