Open Intelligence technology gives you the means to decide with visibility, confidence and speed.
Sourced answer
Le préavis de résiliation est de 90 jours pour les contrats cadres 1 et de 30 jours pour les avenants signés après 2024 2.
They trusted us
OpenOS
OpenOS is neither a model nor a simple agent. It is a set of services, each responsible for one thing, that turns a human-approved goal into durable, observable and verified work.
Every piece of information keeps its source, its date and its confidence level. The search combines words, meaning and the links between pieces of information to find the right passage.
An approved goal becomes a plan broken down into tasks. Each task has its conditions, its expected result and the evidence that will allow it to be closed.
A language model can propose an answer or a plan. Deterministic services verify that the work is authorised, executable and proven.
How OpenOS works
The request is reframed as a goal, with a verifiable success criterion.
The system retrieves what it already knows: documents, past exchanges, procedures.
The goal is broken down into tasks, each with its conditions and its expected result.
The rules and human approvals apply before any execution.
Specialised agents carry out the tasks with the tools made available to them.
A task is only closed if the expected evidence is there. The model’s confidence is not enough.
Each step leaves a trace linked to the others, which can be reviewed afterwards.
Architecture
The point of contact: identity, conversation, memory, search, goals and final answer.
The plans, the rules, the approvals, the task queue and their final control.
The agents that execute: typed adapters, tools, progress tracking.
The record of the work, human-readable, with the links to the evidence.
The audit log: events, traces and review of the evidence.
The contracts between services: identities, signatures and validation of exchanges.
Companion
AI
Collaboration
Weekly summary
New
Summary sheet
Generated by the agentSecurity
Logging of queries, consulted sources and actions.
SSO and account provisioning from your directory.
Answers only expose what the user can already access.
Hosting in Europe, in a private cloud or on your own infrastructure.
Every answer can be replayed with its original sources.
Data encrypted in transit and at rest.
OpenOS is the AI operating system we designed. It brings together a persistent memory, a search that cross-references several signals, step-by-step planning, specialised agents and an evidence log.
Yes. It is designed and maintained by our teams. It works with the language model of your choice and depends on no model provider.
The core of the system (understanding, planning, execution, evidence) works end to end and has been verified with a real model. Deployment at a customer is handled case by case, as part of a project.
Companion is our desktop application: a memory you can read and correct, cited sources, and agents that plan a piece of work then ask for your approval before acting. It is in alpha.
Sourced answers, an engine tailored to your business, agents able to act in your tools and a deployment compatible with your security constraints.
Yes. The architecture handles several million documents and fine-grained access rights.
Yes, through native connectors, an API and an MCP server.
The fastest way to understand Open Intelligence is to see it answer on your own documents. Describe your need: we’ll come back to you with a tailored solution and a costed quote.
