MCP server
Agents discover your concepts and rules, run reasoning and fetch explanations through Model Context Protocol. Access is scoped to what each user can see.
MCP docs →Deterministic Agents at Scale
Combine your data and everything your organization knows into rules a machine can run and a person can verify, on the stack you already run on.
Proven in production
AI has proven it can generate.
It must operate now.
In regulated industries, AI must be reliable, governed and has to integrate seamlessly within existing systems to get to work, fast.
The model is not your differentiator
Every competitor gets the same models, at roughly the same time. All are probabilistic by design. Your advantage is what you put in front of them.
Data is not quite ready
Fragmented systems, inconsistent definitions, and brittle pipelines keep enterprise data inaccessible.
AI costs are out of control
Agents rewrite queries every single time leading to repeat token spend and unreliable results.
Context is only half of the answer.
Context is stored and lost in heads of experts. Revenue in your CRM means annually recurring. Contract value in your payments system includes one-off service fees that never recur. Both are correct, and your team knows which is which.
But, written as text, its just more for an LLM to read, re-interpret, and then guess a bit better.
The missing half is your rules.
Your unique business rules, codified, governed and executable deterministically by humans and agents over your enterprise data in place.
Learn moreNative infrastructure for deterministic systems
Agents don't only retrieve context, but can execute deterministic business rules directly against live enterprise data, wherever it is.
Take the tour on a bigger screen.The interactive product tour is built for laptops and tablets.
Book Whiteboard SessionBuilt for critical workflows where 99% isn't good enough
When a decision has to stand up to a regulator, an auditor, or a board, Prometheux shows the reasoning behind every answer.
100%
Same input. Same output. Every time.
Trust the answer
Deterministic
Results you can rerun, test, and sign off on.
Explainable
See which rules and steps produced each result. No black box.
Governed
Control access and spend across teams and agents.
Run it anywhere
Owned
Your logic survives warehouse moves, team changes, and tool sprawl.
Portable
On your premises or in any cloud. No migrations, no lock-in.
Scalable
Billions of data points in seconds, on compute you already have.
No Lock-In.
100% Portable Infrastructure
Your business never stands still. Neither should your infrastructure. Move data, swap models, change your rules, and deploy on-premises or in the cloud. No long migrations. No starting from scratch.
Connects to anythingRuns anywhereRidiculously faster
Analytics
Graph and BI
Integrations
Back into your apps
Agents
Run workflows deterministically
Models
Define the ontology, read answers, and power analytics, integrations and agents
Your data sources
Connected in place: warehouses, databases, files and APIs. Nothing to migrate.
Agents that can show their work.
Connect any agent to your ontology over MCP or API. Every answer is deterministic and traces back to the exact rows it came from.
risk-desk — claude
✻ Welcome to Claude Code!
cwd: ~/risk-desk · MCP: prometheux connected
cwd: ~/risk-desk · MCP: prometheux connected
> which of our counterparties are controlled by a sanctioned entity?
⏺ prometheux - reason (MCP)(ontology: "ownership", query: "flag(C)")
⎿ 4 rows · deterministic · 1.4s
⏺ 4 counterparties have a sanctioned ultimate controller:
• Nordhavn Capital 3-hop ownership chain
• Brisa Logistica SpA 2-hop ownership chain
• Kestrel Trade Ltd joint control, 2 holders
• Orvelle Metals direct 54% stake
> why is Nordhavn Capital flagged?
⏺ prometheux - explain (MCP)(fact: "flag(Nordhavn Capital)")
⎿ owns(Varko, Eastline BV, 0.62) ← snowflake #88213
owns(Eastline BV, Nordhavn, 0.70) ← postgres #1204
sanctioned(Varko) ← s3 sdn_list.csv:5531
>
? for shortcuts⏵⏵ auto-accept edits on
risk-pipelines — Cursor
ownership.rulesnightly_flags.py
1% Ownership ontology · defined once, reused everywhere 2controls(X,Y) :- owns(X,Y,S), S > 0.5. 3controls(X,Z) :- controls(X,Y), controls(Y,Z). 4flag(C) :- controls(S,C), sanctioned(S). 5 6% sources 7@bind("owns", "snowflake", "kyc.shareholders"). 8@bind("sanctioned", "s3", "sanctions/sdn_list.csv").
AgentNew chat
Add a nightly job that writes flagged counterparties to risk.flags
Called MCP tool prometheux.list_concepts ✓
Called MCP tool prometheux.reason ✓
The flag rule already exists in ownership.rules, so the job only schedules it and writes the output.
nightly_flags.py+18 −0
Plan, search, build anything
∞ AgentAuto↑
chatgpt.com
ChatGPT ⌄
Which suppliers would we lose if Eastline BV were sanctioned tomorrow?
Used Prometheux
Three active suppliers are controlled by Eastline BV, directly or through subsidiaries. Together they make up 18% of Q3 spend.
| Supplier | Control path | Q3 spend |
|---|---|---|
| Nordhavn Capital | Eastline → Nordhavn | €4.2M |
| Halden Components | Eastline → Nordhavn → Halden | €2.9M |
| Vireo Freight | Eastline → Vireo | €1.1M |
+Ask anything↑
risk-desk — px — zsh — 100×30
maya@mbp risk-desk % px concepts ls --ontology ownership
Company Person owns controls sanctioned flag
maya@mbp risk-desk % px reason ownership "flag(C)" --stream --format jsonl
{"C":"Nordhavn Capital","hops":3}
{"C":"Brisa Logistica SpA","hops":2}
{"C":"Kestrel Trade Ltd","hops":2}
{"C":"Orvelle Metals","hops":1}
✔ 4 rows · 2.1M edges scanned · 1.4s · deterministic
maya@mbp risk-desk % px plan ownership "flag(C)"
scan snowflake.kyc.shareholders pushdown
scan postgres.registry.holdings pushdown
recurse controls/2 → join sanctioned/1 → flag/1
maya@mbp risk-desk % px explain "flag(Nordhavn Capital)" --open
↗ derivation opened in Prometheux · link copied
maya@mbp risk-desk %
Any MCP-capable agent
Claude, ChatGPT, Cursor, Copilot, or straight from the terminal
API & SDK
Call the same ontology from services, notebooks and pipelines over REST or Python. One definition serves apps and agents alike.
API docs →>_
CLI
Stream results, inspect query plans and open derivations without leaving the terminal. Coding agents call it directly, with no server to set up.
CLI docs →From first conversation to production‑ready in three weeks.
A fixed plan with clear outcomes each week, so you always know what's happening and when we need your team.
Week 01 Scoping
Outcome: an agreed use case, scope and success criteria.
Monday
Whiteboard
Walk through your problem, current workflow and goals with our human in the loop.
Working sessionTuesday
Build
Prometheux prepares a tailored demonstration unique to your business and workflows.
BuildWednesday
Deep Dive
Explore Prometheux with our engineers and refine requirements.
Working sessionThursday
Review
Your team discusses priorities and chooses the preferred use case.
YouFriday
Use Case Decision
Confirm the scope, success criteria and sample data requirements.
Decision
Week 02 Proof of Concept
Outcome: a working proof of concept tested against your agreed goals.
Monday
Bootcamp
Map your data, workflow, business rules and exceptions with our engineers, and prepare sample data together.
In person preferred. We come to you. Online available.
Working sessionTuesday
Bootcamp
Review the workflow map, captured rules and build plan together, and confirm our understanding.
In person preferred. We come to you. Online available.
Working sessionWednesday
Build
Prometheux connects the sample data and implements the workflow.
BuildThursday
Build
Prometheux tests results, investigates exceptions and refines the logic.
BuildFriday
Business Case Decision
Run the workflow together, inspect the results and validate for production.
Decision
Week 03 Production-ready
Outcome: an initial workflow ready to go live in production, plug-and-play with your existing systems, in your preferred environment.
Reserve whiteboard session →Next availability: within a weekStart where your data already is
Built & Backed By
Where is 99% not good enough?
Bring us one critical workflow. Our Forward Deployed Engineers will help you run it autonomously and deterministically at scale in days, not months, on the data and infrastructure you already have.
Start with one workflow today.
Scale your entire operations tomorrow.
Scale your entire operations tomorrow.
