AI Agents
Autonomous AI workers, inside your perimeter
An agent is a prompt, a model and read-only access to your data. Create it self-service, expose it on the interfaces you enable: it works on your data, with your rules, without ever leaving your perimeter.
Specifications
Definition
A prompt, a model, your data
Data access
Read-only, scoped by you
Models
Local vLLM or European APIs
Governance
Inherited from the platform
Creation
Self-service
Exposure
The interfaces you enable
Use cases
Create a business assistant
An agent that answers your team's questions from your data, created self-service
Created in minutesRun inside your perimeter
Local models served by the platform (vLLM) or European APIs: inference runs where you decide
No data exposedExpose under control
Each agent is only reachable from the interfaces you enable, with read-only access to data
Controlled exposureIn action
The business lead
Nadia wants an assistant that answers her team's questions about business data
- 1 Create an agent from the console: write the prompt
- 2 Pick a model served by the platform
- 3 Define the accessible data scope, read-only
- 4 Enable the access interface for her team
A working business assistant, with no AI project and no ticket
The confident CIO
Marc needs to open AI to business teams without losing control of the data
- 1 Check that every agent inherits the platform's governance
- 2 Control data scopes: read-only, nothing more
- 3 Choose where models run: local or European APIs
- 4 Audit the agents created by the teams from the console
Business teams create their agents, IT keeps control
Key benefits
Ready to create your first AI agent?
Discover how AI agents work on your data without exposing it.
Request a demo