Questions
Short answers to what infrastructure and security teams ask first.
The basics
What is octroi.ai?
A control plane for infrastructure operations. Your team publishes a closed catalog of typed Actions. Members and agents propose Jobs from it, the required people approve them in a Channel, and a dispatcher placed in your network runs them. Every step is an Event.
Is it useful without AI agents?
Yes. Members use the same catalog, the same approvals and the same audit trail. Agents come later, as Principals under the same rules, when you decide.
Which agents and models does it work with?
Any agent that can call a typed API. Each Action carries an input schema that maps directly onto a model's tool definition, so the model stays interchangeable.
Does octroi.ai know what my scripts do?
No, and it does not claim to. It records who authorised which Action, on which Platform and Environment, with which rights. What a script does is the contract your team reviewed when it added the Action to the catalog.
Security
Does the agent hold credentials to my infrastructure?
No. An agent holds a token that lets it propose Jobs within its scope and budget. The credentials that reach your systems belong to the dispatcher, which runs inside your network and exposes no API.
Where are secrets stored?
With your dispatcher, in your own infrastructure, never in octroi.ai. The dispatcher hands them to the Job at run time and masks them in logs and outputs.
Who approves a Job?
The effective requirement is fixed when the Job is proposed: the strictest of the Action, its risk, the Environment, the team and the agent's autonomy level. It is automatic, one human, two distinct people, or a refusal. No layer can lower it.
What stops a manipulated agent from doing damage?
The prompt is not the boundary. An agent can only propose Actions from the catalog, within its scope, and the state machine decides what waits for a human. A manipulated agent can propose; it cannot approve its own Job or step outside the catalog.
Day to day
Can an agent raise its own autonomy level?
No. Only a human raises a level, and the change is recorded as an Event. A failed Job the agent had approved automatically lowers its level by one.
What happens when a run is interrupted?
A Job that cannot be retried safely moves to UNKNOWN and a human is notified. octroi.ai does not guess: a person establishes what happened and resolves the Job.
How do I start?
Open the app, create your organization or accept an invitation, and catalog a first Action. You place the dispatcher in your own network.