Under the hood · For developers

Connect your AI, whichever it is. The ontology sets the rules.

The model never sees the database. It sees tools with business names, and only the ones its token allows it to use.

Architecture

Five building blocks, one cycle that repeats.

  1. 01 · Ingestion

    ERP, eCommerce, CRM, CSV or APIs, with no manual pipelines.

    • ERP
    • CRM
    • eCommerce
    • CSV and PDF
    • API
  2. 02 · Data lake

    Versioning and lineage built in, partitioned by customer and by time.

    • Versioning
    • Lineage
    • Partitioning
  3. 03 · Ontology

    Objects, relationships and actions, readable by AI via MCP.

    • Objects
    • Properties
    • Relationships
    • Actions
    • Permissions
  4. 04 · Permissions per token

    Applied before answering: role, rows, columns and actions.

    • Role
    • Rows
    • Columns
    • Actions
    • Audit
  5. 05 · Hypotheses

    Correlations detected and validated against real data.

    • Patterns
    • Root cause
    • Anomalies
    • Alerts

MCP server

Your company, inside your AI.

Installation

Connect KnowOS to Claude, ChatGPT or any MCP client.

One endpoint and the same contract for all of them. You switch models without touching the integration.

Tools

Business names, not tables. Reading needs no approval; acting always does.

API

REST and GraphQL generated from the data models.

Audit

Who asked, with which token, which tool and what they received.

Relational + vector

What happened and what was said, anchored to the same object.

  • Exact SQL over ERP, CRM and WMS
  • Search by meaning across price lists, manuals and email
  • Call audio transcribed and chunked
  • One question spans both worlds and cites both sources

Technical questions

What APIs does KnowOS offer?

REST and GraphQL APIs generated from the data models, and an MCP server for AI clients. The same contract for all of them.

Do I need a data engineer?

No. Ingestion, the data lake, modelling and the ontology are configured without code during setup.

Which AI model does it use?

Whichever you prefer: GPT, Claude or Gemini. You switch models without touching the integration, because the rules live in the ontology.

Does it write to my ERP?

Not in the first phase. After that, actions are generated as drafts and Connect or Flow executes them once a person approves.

Is there an on-premise option?

On request. By default, a dedicated EU server in your name.

Show us a question that takes an afternoon today.

A 30-minute session on a real case: a historical price list, an internal procedure, a query that today ends up as a request to IT. You leave with the scope of sources and a fixed price.