# Prometheux > Prometheux is the knowledge graph platform built for on-prem AI. It queries data across databases, warehouses and platforms as one knowledge graph, wherever the data lives, and infers new knowledge from it with business rules. No migrations, no graph database to run, and every answer is deterministic and explainable. ## What Prometheux is Prometheux is a knowledge graph platform with its own reasoning engine. It is not a graph database: you query your existing data as a graph without building one or copying it into a separate store. - **Query any data as a knowledge graph.** Prometheux connects to SQL databases, NoSQL stores, data lakes, files and REST APIs and processes data where it lives. Tables can be queried as graphs, graphs as tables, or any combination, in one place. - **Infer new knowledge.** Business rules derive what no table stores, such as which supplier a plant depends on three tiers down or the total exposure across a chain of entities. The engine handles recursion and recursion with aggregation, so multi-hop and network questions run as rules. - **Write in any language.** Natural language, SQL, Cypher, Python, Vadalog and Datalog, all in one engine. You or an agent define concepts, relationships and logic once, and they run across all connected data. - **Scale out.** The distributed engine parallelises graph traversal and data processing across the machines and clusters you already have, from sub-second queries to petabyte-scale jobs. - **Explainable by design.** The same question always returns the same answer, and every result can be followed back through the concepts and sources it depends on. - **Ready for AI agents.** Any model (Claude, GPT, Gemini, open-source) uses Prometheux through MCP, the API or the CLI, so agents query the knowledge graph and get answers they can explain instead of guessing from raw tables. ## Why not a graph database or a semantic layer? - A graph database needs its own copy of your data, kept in sync, plus a migration and a second data stack. Prometheux reads the data where it is. - Most graph platforms run one query on one machine because a graph is hard to split across machines; scaling means buying a bigger server. Prometheux distributes the work across a cluster. - Graph tools only handle nodes and edges, so the rest of the workflow needs other tools. Prometheux runs the whole pipeline (connect, model, reason, serve) in one engine. - A semantic layer gives metrics consistent definitions. Prometheux also executes business rules over the data and derives new facts, with an explanation for each one. ## Where it runs On your premises, in your own AWS, Azure or GCP account, as a native app on Snowflake or Databricks, on Palantir, or as SaaS fully managed by Prometheux. Knowledge graphs stay portable, with no lock-in to a database, cloud or AI model. ## Security SOC 2 Type II audited (report shared on request). Deploy where your data has to stay, including fully on-prem. Each user's storage is isolated, and agents act with the user's own access rights. ## Proven in production - Dompé: 30+ sources queried in one unified knowledge graph - AstraZeneca: 9x faster graph reasoning (with NVIDIA) - Banca Sella: under 1 minute per regulatory credit-risk flow, with explanations - A Fortune 100 retailer: +42% retrieval improvement - A financial institution: €9M saved per year Also trusted by the Central Bank of Italy, Vodafone and Novogaia. Working with the University of Oxford, TU Wien, Roma Tre University and the University of Cambridge. ## Common questions - **Is Prometheux a graph database?** No. It's a knowledge graph platform with its own reasoning engine that queries your existing data as a graph. - **Do I have to move my data?** No. It processes data where it lives: no migration, no duplicate copy to keep in sync. - **What is a knowledge graph without a graph database?** A graph view over the tables, files and APIs you already have, defined by the concepts and relationships of your business, queried and reasoned over in place. - **How fast can I start?** Connect a source, define the concepts that matter (or let the assistant draft them), write rules, run them, and use the results from your apps, dashboards or any MCP agent, in under five minutes. Prometheux engineers can help with the first rules. ## Key pages - [Home](https://www.prometheux.ai/): The knowledge graph platform built for on-prem AI - [Platform](https://www.prometheux.ai/platform): How it works, architecture, deployment and security - [Knowledge Graph](https://www.prometheux.ai/capabilities/knowledge-graph-builder): Query any data as a graph, nothing to build - [Virtual Knowledge Graph](https://www.prometheux.ai/capabilities/virtual-knowledge-graph): Query a graph over data that never moves - [Graph Analytics without a Graph Database](https://www.prometheux.ai/capabilities/graph-analytics-without-a-graph-database): Query your tables as a graph, no ETL - [High-Performance Data Engine](https://www.prometheux.ai/capabilities/high-performance-data-engine): Distributed reasoning on clusters and GPUs - [Vadalog Playground](https://www.prometheux.ai/playground): Write and run Vadalog in the browser, no sign-up - [Documentation](https://docs.prometheux.ai/): Guides, the Vadalog language reference, SDKs and the API - [Partners](https://www.prometheux.ai/partners): Databricks, Snowflake, NVIDIA, AWS - [Resource Center](https://www.prometheux.ai/resource-center) and [Research](https://www.prometheux.ai/research) - [Book a Demo](https://www.prometheux.ai/book-a-demo) and [Contact](https://www.prometheux.ai/contact) ## Capabilities - [Context as Code](https://www.prometheux.ai/capabilities/context-as-code): Policies and notes as governed, executable context - [Domain-Aware AI](https://www.prometheux.ai/capabilities/domain-aware-ai): Ground LLMs in your business rules - [Knowledge Graph](https://www.prometheux.ai/capabilities/knowledge-graph-builder): Query any data as a graph, nothing to build - [Ontology Generation](https://www.prometheux.ai/capabilities/ontology-generation): AI drafts executable ontologies from plain language - [Semantic Layer](https://www.prometheux.ai/capabilities/semantic-layer): Shared business definitions that actually execute - [Virtual Knowledge Graph](https://www.prometheux.ai/capabilities/virtual-knowledge-graph): Query a graph over data that never moves - [AI Lineage](https://www.prometheux.ai/capabilities/ai-lineage): Trace AI answers back to source data - [Data Lineage](https://www.prometheux.ai/capabilities/data-lineage): Row-level provenance across every source - [Deterministic AI](https://www.prometheux.ai/capabilities/deterministic-ai): Same input, same answer, every time - [Natural Language to Query](https://www.prometheux.ai/capabilities/natural-language-to-query): Ask in English, run as logic - [Neurosymbolic AI](https://www.prometheux.ai/capabilities/neurosymbolic-ai): LLM flexibility, logic you can audit - [Token Spending Control](https://www.prometheux.ai/capabilities/token-spend-control): Deterministic engine work, fewer LLM tokens - [AI Governance](https://www.prometheux.ai/capabilities/ai-governance): Business policy as executable, auditable rules - [AI Observability](https://www.prometheux.ai/capabilities/ai-observability): See every AI step, input and result - [Cross-Database Processing](https://www.prometheux.ai/capabilities/cross-database-processing): One program across every database - [Entity Resolution](https://www.prometheux.ai/capabilities/entity-resolution): Explainable record matching across every source - [Graph Analytics without a Graph Database](https://www.prometheux.ai/capabilities/graph-analytics-without-a-graph-database): Query your tables as a graph, no ETL - [Graph Traversal](https://www.prometheux.ai/capabilities/graph-traversal): Follow relationships at any depth - [Graph Visualization](https://www.prometheux.ai/capabilities/graph-visualization): See your ontology and trace its lineage - [High-Performance Data Engine](https://www.prometheux.ai/capabilities/high-performance-data-engine): Distributed reasoning on clusters and GPUs - [Recursion](https://www.prometheux.ai/capabilities/recursion): Hierarchies, supply chains and BOMs as rules - [Fast Migrations](https://www.prometheux.ai/capabilities/fast-migrations): Move data without rewriting logic - [Faster ETL](https://www.prometheux.ai/capabilities/faster-etl): Pipelines as a few readable rules - [In-Place Data Processing](https://www.prometheux.ai/capabilities/in-place-data-processing): Move the logic, not the data - [Portable Cloud](https://www.prometheux.ai/capabilities/portable-cloud): Any cloud, on-prem, Databricks, Snowflake - [Elastic Compute](https://www.prometheux.ai/capabilities/elastic-compute): Right-sized machines, on when needed - [On-Premises Deployment](https://www.prometheux.ai/capabilities/on-premises): Sovereign AI inside your own data centre - [Agentic Workflows](https://www.prometheux.ai/capabilities/agentic-workflows): Agents that call your governed logic - [Vertical AI Training](https://www.prometheux.ai/capabilities/vertical-ai-training): Fine-tune on verified, reasoning-generated data ## By Industry - [Finance & Banking](https://www.prometheux.ai/industries/finance-banking): Every financial decision, with the evidence - [Insurance](https://www.prometheux.ai/industries/insurance): Find fraud rings, pay honest claims - [Pharma & Life Sciences](https://www.prometheux.ai/industries/pharma): Discovery and quality, every step traced - [Retail](https://www.prometheux.ai/industries/retail): Recommendations with a reason - [Manufacturing & Ops](https://www.prometheux.ai/industries/manufacturing): Every failed batch traced to its cause - [Energy & Utilities](https://www.prometheux.ai/industries/energy): Outage impact and reports you can prove - [Telecommunications](https://www.prometheux.ai/industries/telecommunications): The knowledge plane for autonomous networks - [Public Sector & Government](https://www.prometheux.ai/industries/public-sector): Explainable decisions on sovereign infrastructure - [Supply Chain & Logistics](https://www.prometheux.ai/industries/supply-chain): Multi-tier supplier risk, traced end to end - [Other Industries](https://www.prometheux.ai/industries/other-industries): Industry agnostic: your rules, your data ## By Role - [For Agents](https://www.prometheux.ai/for/agents): Business rules as tools for agents - [For Data Engineers](https://www.prometheux.ai/for/data-engineers): Declarative pipelines across every source - [For Executives](https://www.prometheux.ai/for/executives): AI you can defend, logic you own - [For Startups](https://www.prometheux.ai/for/startups): Ship agents, skip the data plumbing ## Use Cases - [AI Agents](https://www.prometheux.ai/use-cases/ai-agents): Take AI agents from pilot to production - [AI Cost Control](https://www.prometheux.ai/use-cases/ai-cost-control): Pay for language, not for logic - [Data Modernization](https://www.prometheux.ai/use-cases/data-modernization): Modernize without a big-bang migration