Onboarding
Live in five steps — guided by the AI agent.
Auto-onboarding means you don't wait on us. The business agent guides the cadrage; the platform provisions everything else.
-
Create your account
Sign up, and your tenant is provisioned automatically with RBAC roles (admin, business, data scientist).
-
Install the data plane
Run one command in your environment. The Runner self-registers over an outbound connection — no inbound port.
-
Connect your data
Declare your sources in a wizard. Credentials stay inside the Runner; only schemas get profiled.
-
Use the AI agents
The business and data-science agents map your schema to a use-case template and generate the code.
-
Run your process
Execute the scenario near the data. Scores and metrics flow back into monitoring dashboards.
Create your account & tenant
You sign up (or accept an invitation). A tenant is provisioned automatically in the control plane with RBAC roles — tenant admin, business, data scientist. No ticket, no waiting on us.
- Self-serve sign-up
- Automatic tenant provisioning
- Role-based access from the start
Install the data plane Runner
The SaaS issues a single-use, short-lived, tenant-scoped enrollment token. You run one turnkey command in your environment; the Runner self-registers over its outbound connection and shows up as “online”.
- One command: docker run kepler-runner --enroll <token>
- Outbound self-registration — no inbound port
- Token is single-use, short-lived and revocable
Connect your data sources
A wizard collects host, port, credentials and paths. Those credentials stay inside the Runner — the SaaS only keeps a connection reference. The Runner tests the connection and profiles the schema.
- Credentials never leave the Runner
- Connection tested (conncheck) before use
- Schemas profiled — agents see metadata, not rows
Use the AI agents
Pick a use-case type and import a template. The business agent frames the problem; the data-science agent maps the template’s expected features to your real schema and generates the transform, training and scoring code.
- Typed use cases: KYC, AML, fraud, credit, IFRS
- Template features mapped to your schema
- Code generated from metadata only
Run your process
The scenario executes in your data plane — training and scoring run near the data. Aggregated results and metrics flow back into monitoring dashboards with explainability and drift tracking.
- Near-data execution in a sandbox
- Real-time scoring served from the Runner
- Results, SHAP and drift in monitoring