01 / Product Detail
Structure data inquiries into Data MCP, Data Agent, and Data App components.
02 / Artifacts
KunimiAI produces three concrete artifacts.
01
Data MCP
Governed interfaces for data access
Datasets, metrics, policies, and lineage are exposed as MCP tools. Access is defined by the applicable catalog permissions and policies.
Evaluation lens: can existing metrics, dbt assets, and policies be called safely as MCP tools?
02
Data Agent
Conversational agents that translate queries into executable workflows
Natural-language inquiries are compiled into structured analysis units tracking referenced data, generated code, intermediate tables, and review history.
Evaluation lens: can the same ask be rerun by another team, another period, or another data snapshot?
03
Data App
Operational interfaces for analyzed data
Analysis outputs can be published to operational workflows, APIs, MCP endpoints, BI dashboards, or spreadsheets.
Evaluation lens: where does the output connect to board review, weekly review, or daily operation?
03 / Product Flow
A data workflow progresses through five stages.
- 01
Ask
A user asks in business language. Example: why did churn rise among our key accounts?
- 02
Context
KunimiAI resolves metric definitions, governance policies, and catalog permissions to establish the query scope prior to execution.
- 03
Execute
The agent generates SQL, Python, or workflow logic and runs it against Iceberg snapshots and semantic assets.
- 04
Review
Analysts, domain experts, and data owners inspect the answer, generated code, source references, and approval state in one place.
- 05
Promote
Only valuable work is promoted into a Data MCP, Data Agent, or Data App for reuse.
04 / Evaluators
Each reader should know what to inspect.
CDO / Head of Data
Evaluate how Data MCP, Data Agent, and Data App components align with organizational governance and operational boundaries.
Data Platform Lead
Checks where Polaris, Iceberg, dbt, identity, and audit logs connect.
Analyst / BI Lead
Verify that ad-hoc analyses can be converted into version-controlled, reproducible workflows.
Security / Governance
Checks pre-execution denial, post-execution lineage, approvals, and audit logs.
05 / Next Step