NatusAIScoped tools
Each tool has a clear effect. Separate permissions decide what an assistant may do with it.
AI coordination · Layer 02 · IntelligenceRoadmap
NatusAI Auriga: AI assistants that work within the limits you set
Auriga is planned as the coordination layer for AI assistants and tools. Each task gets the context it needs, works within permissions and budgets set by people, and stays a proposal until it passes checks and is accepted.
- Routine work handled within agreed boundaries; exceptions go to a named person.
- A proposal is not permission: nothing consequential happens without approval.
- Cost and quality measured together: completion, retries, time and total spend.
What it does for the business
Give each task the context it needs, and nothing more
Most AI risk comes from assistants that see too much, act too freely or cannot show their work. Auriga is designed around four rules.
Relevant context
Only the documents and data a task needs, with their source and freshness visible.
Independent checks
AI work stays a candidate until it passes its checks and a person or approved process accepts it.
Measured efficiency
Success judged together with retries, time and total cost. No token-saving claim is made here.
Retrieved documents and model output are treated as content to inspect, never as instructions that grant authority.
Honest status
What you can use today, and what is on the roadmap
We separate what is proven from what is coming, and label both. NatusAI Auriga is on our roadmap.
What you can use today
- Delivery engagements that automate routine steps with human handoff, as in the telecom result below.
- Assistants that answer from approved sources and route complex cases to people.
- Model choice under your policy: hosted, EU-hosted or running locally.
What’s on the roadmap
Roadmap- Auriga as a product: durable tasks with checkpoints, cancellation and budgets.
- Tool discovery under each caller’s permissions.
- Independent verification before any result is promoted. Auriga is not part of the planned Cygnus first release.
How it connects
Auriga coordinates AI work across the platform: it reads from NatusAI Cygnus and Aquila, presents through NatusAI Lyra, and its proposals become reviewed releases in NatusAI Orion.
For your technical team
- Tools expose typed effects; permissions are granted separately and checked at execution.
- Agents discover capabilities and retrieve versioned context (schemas, documentation, source, evidence) with source and freshness visible, and return structured results for review.
- Durable attempts, checkpoints, cancellation and budgets, so work and its evidence survive beyond one prompt or session.
- Retrieved documents, database text and model output are treated as untrusted content, never as instructions that grant authority (prompt-injection posture).
- Provider and capability support is qualified for each exact integration. Broad provider-native agents, background collaboration and automatic fallback are not part of the planned Cygnus 1.0.
- Planned Cygnus 1.0 includes a durable human decision and approval queue, and native MCP descriptors whose discovery and execution follow the caller’s permissions.
Control by design
Sovereignty, governance and security at NatusAI How NatusAI Auriga de-risks AI
Ownership
Model-agnostic by design. Each AI provider integration is qualified, not assumed.
Evidence
Results return in a structured form with their sources, ready for review.
Human review
Approvals stay with people. An assistant cannot widen its own access.
Security
Text an assistant reads can never instruct it to act. Authority comes only from your permissions.
Related delivered work
Where it is used
Where our team has applied this capability on the engine behind our client projects, and where it is shown in live demos or reference scenarios. NatusAI Auriga is how we are turning it into a product.
Delivered · client result
Telecom
A European telecommunications operator
92%
faster response time
Read the case
Reference scenario
Use case
Customer Support
Answer customer emails from approved knowledge. Routine tasks complete within agreed limits; exceptions go to your team.
Explore the use case
Capability pattern
Use case
Knowledge Graphs for AI answers
AI answers grounded in your facts and their relationships, traceable to the source and correct for the date asked.
Explore the use case
Works with
Layer 02 · IntelligenceRoadmap
NatusAI Aquila
Turn documents into knowledge you can cite, with the source one click away.
Explore NatusAI Aquila
Layer 01 · FoundationPlanned v1
NatusAI Cygnus
Relationships, history, location and search in one connected picture of your business.
Explore NatusAI Cygnus
Layer 03 · ApplicationsRoadmap
NatusAI Orion
Describe the application you need, review the result, release it with approval, then run and improve it.
Explore NatusAI Orion
Questions
Questions about NatusAI Auriga
Short answers for decision-makers. Your technical team will find the detail in the disclosure above.
Can an AI assistant act without approval?
Only inside boundaries you set in advance, such as answering a routine enquiry from approved sources. Anything consequential waits for a named person.
Which AI models does Auriga work with?
The ones your policy allows: hosted, EU-hosted or running locally. Each provider is qualified for the exact integration before use.
Is NatusAI Auriga available today?
Not as a product. Today our team delivers coordinated AI with human handoff inside client engagements.
Talk to our team
Where could coordinated AI work help?
Bring a task, the people responsible for it and the rules it must follow. We will show you what can be automated safely.
Or write to contact@natusai.com
