Ship Agents.
Skip the
Plumbing.
Your stack, cloud and models will change. Your business logic shouldn't have to be rebuilt each time. Write it once as rules that run where your data lives, give your agents answers they can repeat, and keep the same rules as you scale.
Proven in production
Build Fast.
Not Twice.
Early AI products often bake the business logic into prompts, glue code and a data pipeline built for the demo. It works until the first big customer asks how an answer was reached. Start with logic you can reuse, run and explain.
Rules In.
Agents Out.
Write the logic your product depends on as rules in Vadalog, SQL or Python, over the data you already have. Your agents and apps call those rules through MCP, the REST API or the Python SDK, and get a computed answer with its trail attached.
Agent-Ready From Day One
A remote MCP server with OAuth works with Claude and other MCP clients. REST and Python cover the rest.
No Data Platform to Build
Rules read your warehouses, buckets, databases and files where they are. Nothing to migrate first.
Answers You Can Repeat
Same data, same result. LLM steps are labelled, so you know which parts of a run could vary.
A Trail Your Customers Trust
Each result links back to the rule and source rows that produced it, ready for an enterprise buyer's review.
Ship the feature, not the pipeline. Show enterprise buyers how every answer was reached. Keep your logic yours as you scale.
Small Today.
Large Tomorrow.
Develop on a small machine, run production on a large one, and turn it off when the work is done. Repeatable logic runs once in the engine, so LLM calls are kept for the language work.
Right-Sized Machines
Pick a machine per job from the catalog, and set an idle timeout so nothing runs when no one is using it.
Same Rules, Any Size
Every Prometheux-managed machine reads the same workspace, so switching to a bigger machine is a click, and no data moves.
Fewer, Smaller LLM Calls
Joins, filters and multi-step logic run in the engine. The prompt gets the result, not the raw tables.
Portable From the Start
Start on managed SaaS, then run in your own AWS, Azure or GCP account, on-premises or natively on Databricks, with the same rules.
Pay for compute while it works, not while it idles. Keep tokens for language, not logic. Move clouds or models without starting over.
Connect. Define.
Run.
Connect
Point Prometheux at the data your product uses: Snowflake, Databricks, S3, Postgres, files. Nothing is copied.
Define
Write your product's logic as rules, or bring the SQL you already have. Version it like code.
Run
Your agents and app call the rules through MCP, the REST API or the Python SDK, on a machine sized for the job.
Grow
Move to a bigger machine or a new cloud. The rules stay the same.
Products That
Need Answers
Deterministic Agents
Give your agent rules to call instead of logic to guess, so it answers the same way every time.
Deterministic AI →Agentic Workflows
Chain rules, LLM steps and schedules into workflows your customers can run and inspect.
Agentic workflows →Token Spend Control
Run repeatable logic once in the engine and keep LLM calls for language, with usage and limits you can see.
Token spend control →Deploy Where They Are
When an enterprise buyer asks how an answer was reached, every result shows the rules and sources behind it.
Portable cloud →“Prometheux enabled us to build a system that can access our databases, process all the client's historical variables, and provide an intelligent, understandable explanation autonomously.”
Massimiliano MalinHead of Credit Risk · Banca Sella
“Prometheux technology powers highly scalable and explainable reasoning over some of the world's largest knowledge graphs.”
Accelerating Neurosymbolic AI with RAPIDS and Vadalog Parallel
Questions,
Answered
Do we need a data platform before using Prometheux?
No. Prometheux connects to the databases, warehouses, buckets and files you already have and runs rules on that data where it lives. There is no mandatory migration.
What does compute cost us when nothing is running?
You choose a machine per job and stop it when the work is done, or let an idle timeout stop it for you. Your rules, files and results stay in your workspace between runs.
Will it hold up when we land a big customer?
The same engine runs regulatory flows at Banca Sella and reasoning over knowledge graphs with billions of edges. Every answer shows the rules and sources behind it, and Prometheux is SOC 2 Type II certified.
Will we be locked in as we grow?
No. Rules are readable text you keep in your own version control, the same rules run on SaaS, your VPC, on-premises or natively on Databricks, and you choose the LLM provider and model per step.
Can our existing agents use it?
Yes. Any MCP client can connect through the remote MCP server with OAuth, including Claude. Anything else can use the REST API or the Python SDK.
Prometheux For…
Building Something?
Talk to Us.
Tell us what you're shipping and the data behind it. We'll show you your logic running as rules your agents can call.
