appendix i — questions

Frequently asked questions.

How the platform works, how agents touch your systems, and where everything runs. Something missing? Send word through the contact page.

What is Obsidura?
Obsidura is an enterprise AI agent orchestration platform. Agents connect to your systems of record - databases, APIs, and business applications - execute routine operational work, and escalate to a human only when judgment is required.
What is Pantheon?
Pantheon is our orchestration suite: typed connectors into your backend, a planner that decomposes jobs into steps, a durable runtime that survives crashes and slow upstreams, and an append-only audit log covering every action.
How do agents access our systems?
Through typed connectors - Postgres, REST, gRPC, and message queues - using scoped, audited credentials. Credentials are minted per step with least privilege and revoked on completion.
What happens when an agent is not confident?
When confidence drops below your threshold, the agent stops and escalates to a human queue with the full decision trace attached, so a person can resolve it in seconds instead of re-deriving context.
Is every action logged?
Yes. Every action lands in an append-only audit log with the full prompt, tool call, and resulting diff. Any run can be replayed at any time.
Where can Obsidura run?
Three deployment options: our fully managed cloud, a private VPC inside your own AWS or GCP account, or air-gapped on-premises on your hardware.
How do we get started?
Book a demo through the contact page. We map one of your workflows on a 30-minute call and show you the audit log by the end of it. We are currently onboarding design partners.