Open-source MLOps you own, with the team that built it on call.
FRAKMA-X has three layers. An Apache-2.0 Community Edition you can run yourself. GitOps workflows that take models from data to production. And an enterprise edition where Warble deploys and supports the whole stack on your cluster. Your cluster, your data, your agents.
Six open-source tools don't wire themselves together
ML pipelines get stitched together by hand and are hard to reproduce. Getting six open-source tools to work together in production takes platform skills most teams don't have. The managed alternatives fix that by moving your data path and models into a vendor's account and roadmap.
One product, three ways in
Community Edition
The FRAKMA stack under Apache-2.0: OpenFaaS, KubeRay, Argo CD, KServe, MLflow and Airflow, plus the WarbleApp CRD operator (warble.io/v1alpha1), cert-manager, NGINX Ingress, Prometheus, Grafana, OpenTofu and Ollama on GPU spot nodes.
A Linux Foundation wing is forming; no foundation has accepted the project. FRAKMA-X is not an LF or CNCF project.
GitOps MLOps workflows
Data → train → register → serve → monitor, managed from Git. Open source to run yourself; set up by Warble in the enterprise tiers.
Enterprise edition
FRAKMA deployed, configured and supported on your AKS or GKE cluster. Starter $5,000/mo · Growth $15,000/mo · Enterprise custom.
From data to a served model, managed from Git
Bootstrap GitOps
Argo CD app-of-apps wired to your Git repo; every component declared in Git.
Pipelines
Airflow DAGs (GitSync) for ingestion, training and promotion.
Train and tune
KubeRay jobs, including fine-tuning pipelines.
Register
MLflow tracking with artifacts in Azure Blob, GCS or S3.
Serve
KServe InferenceService with canary rollouts.
Monitor and retrain
Prometheus and Grafana dashboards; drift triggers the Airflow retrain DAG.
Switch LLM providers
Each engine variant (Vertex, Groq, Ollama) is a separate WarbleApp manifest: one kubectl apply.
Plans
| Starter | GrowthMOST POPULAR | Enterprise | |
|---|---|---|---|
| PRICE | $5,000/mo | $15,000/mo | Custom |
| TERMS | Billed monthly · 30 days' notice to cancel | Billed monthly · 3-month minimum | Annual contract · custom terms |
| PLATFORM | FRAKMA deployment and configuration on your AKS or GKE cluster | Everything in Starter | Everything in Growth; dedicated AKS/GKE cluster or on-premises; multi-cluster and multi-region |
| AGENTS | Up to 3 WarbleApp agent deployments | Unlimited WarbleApp agent deployments; Reflexion engine deployment (closed beta, shadow mode on by default) and LLM provider config | Private Helm chart registry and operator builds |
| PIPELINES | Airflow DAG setup (up to 5 DAGs); MLflow tracking with Azure Blob artifacts | Custom Airflow DAG development (up to 15 DAGs); KubeRay fine-tuning pipeline; KServe canary rollout config | — |
| PLATFORM OPS | Wildcard TLS, NGINX ingress, cert-manager; Prometheus and Grafana dashboards | — | SOC 2 / HIPAA / GDPR advisory support |
| SUPPORT | Monthly architecture review call; email support | Dedicated Slack channel; bi-weekly architecture and roadmap sessions | Custom SLA; platform engineer embed; quarterly business reviews; training and enablement |
| Start with Starter → | Start with Growth → | Talk to Avinash → |
Every plan includes cluster architecture design, GitOps pipeline setup, LLM agent deployment, an observability stack, secrets and TLS management, and an image registry with CI/CD. No infrastructure markup: you pay your cloud provider directly. Everything is handed back, documented, in your Git repo, and FRAKMA stays Apache-2.0, so you can cancel and run it yourself.
Questions
Is FRAKMA-X open source?
The Community Edition (the FRAKMA stack) is Apache-2.0. The enterprise edition sells deployment, operational support, and custom agent and DAG work on top of it. You can cancel and run everything yourself.
Is FRAKMA-X a Linux Foundation or CNCF project?
No. A Linux Foundation wing is forming, and no foundation has accepted the project. warblecloud, LLC is a Silver Member of the Linux Foundation and the Agentic AI Foundation (AAIF). Membership is not a certification of Warble products.
Which clouds does it run on?
Azure (AKS) and GCP (GKE). AWS is in testing. On-premises is available on the Enterprise tier.
Can it run alongside Kubeflow or SageMaker?
FRAKMA is designed to run alongside them: it is namespace-scoped, and the WarbleApp CRD uses its own API group (warble.io).
Can we use our existing PostgreSQL and artifact store?
Yes. Airflow and MLflow can point at an external PostgreSQL, and artifacts can live in Azure Blob, GCS or S3.
Where does Reflexion fit?
The Growth tier includes Reflexion engine deployment. Reflexion is in closed beta with shadow mode on by default: it proposes fixes and a human approves.
How does this relate to Murmuration?
Murmuration is one-time advisory from $500. The FRAKMA-X enterprise edition is a monthly managed platform.