One Definition.
Same Answer.
Revenue means one thing in finance and another in sales, and your AI agents guess between them. Prometheux turns each definition, rules and hierarchies included, into rules that run on your data where it lives. Dashboards, apps and agents get the same answer, traced to its rules and source rows.
| customer | region |
|---|---|
| Northwind | EU |
| Harbor & Co | UK |
| Bluepeak | US |
Proven in production
Layers Describe.
Engines Execute.
A semantic layer, metrics layer or catalog describes your business: it maps terms to columns, rolls up metrics and documents what things mean. Prometheux executes it. The definition carries the business rules, runs deterministically on data in place across sources, and returns every answer with its lineage.
Context Is Code,
Not Text
Glossaries, table descriptions and metric docs are a dictionary: more text for people and models to read, reinterpret and guess from. Prometheux is the engine. Each business definition is code connected to live sources, so changing it once changes the answer everywhere. Same input, same output.
Write It Your Way
Describe a definition in plain English and the AI assistant drafts the rule, or write it in SQL, Vadalog, Cypher or Python. Experts, analysts and engineers share one set.
Decoupled From Schemas
Consumers ask for business terms, not tables, so source changes don't ripple into every report.
Rules, Not Labels
Definitions hold the logic and they actually run: thresholds, exclusions, hierarchies at any depth. Nobody, human or agent, writes and rewrites hundreds of lines of SQL to use them.
Across Every Source
One definition can join warehouses, databases and files in a single rule, with data left in place.
End the "which number is right?" meeting. Change a definition once, everywhere. Show auditors the logic behind every KPI.
Same Answer for
People and Agents
Agents need more than context to read. Ask whether a plant is at risk under your supplier policy and the agent runs the policy over the data, instead of reading a paragraph about it and turning that into SQL. Prometheux serves every rule and concept through APIs, SDKs, apps and MCP.
Apps and Widgets
Map definitions to tables, charts and cards in Prometheux apps, or bring your own frontend.
APIs and Python SDK
Run definitions from notebooks, services and pipelines through the REST API and Python SDK.
Agents via MCP
Agents call your rules and concepts through MCP or the REST API instead of writing their own SQL. See it with Snowflake Intelligence.
Inside Your Platform
Run on premises, in your cloud or on Databricks and Snowflake, under the access controls you already have.
Ground every agent in your official definitions. Give every team and tool the same answer. Keep the logic in your stack, not a vendor's.
From Glossary to
Running Logic
Use What You Have
Your glossary, catalog entries, metric definitions or the SQL teams already run. Nothing gets ripped out.
Encode the Rules
Write each definition in SQL or Vadalog, rules included, or ask the AI Assistant to draft it.
Connect to Sources
Point definitions at the warehouses and databases that hold the data. Nothing moves.
Serve Everywhere
Expose them to apps, notebooks, services and agents through one set of APIs, with lineage on every result.
One Language Across
Fragmented Systems
Regulatory Metrics
Compute exposure, capital and risk measures from one governed definition, with lineage for the regulator.
Customer and Revenue
Give marketing, finance and sales the same definition of an active customer and of revenue.
Agent-Ready Logic
Agents answer business questions by calling rules and concepts, not by guessing joins from a schema.
Risk Across Tiers
Define "plant at risk" once: any supplier in its chain, at any tier, is disrupted. Every team gets the same list.
“Generative AI without ontologies is an engine without a road. Prometheux builds the road.”
Marco RomeiDeputy CIO · Banca Sella
“The ability to query heterogeneous databases through a flexible ontology makes it possible to infer deep knowledge and automate complex tasks, delivering trustworthy insights that far exceed the capabilities of traditional systems.”
Anna FavaFormer Senior Manager, Data & AI · Dompé
Questions,
Answered
What is a semantic layer?
It translates business terms such as revenue or active customer into queries over your data, so every tool uses the same definitions. Prometheux goes a step further: each definition also carries the business rules and runs on your data.
How is this different from a semantic layer, metrics layer or catalog?
Those tools describe your business: they map terms to columns, roll up metrics and document meaning. Prometheux executes it. A definition can hold thresholds, exclusions and hierarchies of any depth through recursion with aggregation, runs deterministically across sources, and returns results with lineage to its rules and source rows.
Do we have to replace our catalog, semantic layer or BI tools?
No. Keep the catalog and the metric definitions you already trust. Prometheux adds the definitions that need rules and many hops, and serves them to apps, notebooks and agents.
Do I need to move data into Prometheux?
No. Definitions are connected to your existing warehouses, databases and files, and the logic runs where the data lives, on premises, in your cloud or on Databricks and Snowflake.
Is our model locked into Prometheux?
No. Definitions are readable SQL or Vadalog you can keep in your own version control, and an ontology exports to a portable file with its data and metadata. Prometheux runs on premises, in your cloud or on Databricks and Snowflake.
What about query cost and speed?
Definitions run where the data lives. Concepts can return cached results when their inputs haven't changed, so repeat questions don't repeat the work.
How do AI agents use it?
Agents call definitions through MCP or the REST API. They get answers computed from your official logic instead of writing their own SQL, and each answer traces back to the rules and data behind it.
Can analysts write definitions in SQL?
Yes. Definitions can be written in SQL, Vadalog, Cypher or Python, and they can build on each other. Under the hood each one is a concept in your ontology.
Related Capabilities
Define It Once.
Use It Everywhere.
Bring the metrics your teams argue about. We'll turn them into concepts every tool and agent can run.
