Skip to main content

​
First Agent Command

​
Audience

This page is for users approaching AiOS through agent flows, APIs, or a future CLI for the first time.

​
Goal and Scope

This page helps you understand how a first command will work and what is executable today:
  1. CLI/Agent is not a separate control plane
  2. commands still flow through the same server-side authorization chain
  3. the repository does not yet implement those CLI subcommands

​
Core Concepts

  • CLI/Agent commands ultimately invoke the server-side API
  • high-risk production mutations still enter the approval chain by default
  • error semantics, resource model, and task model are shared with the Web channel

​
Standard Workflow

  1. start the current bootstrap/server entrypoint
  2. inspect implemented API status surfaces and error semantics
  3. interpret future CLI commands as wrappers over the same server-side execution chain

​
Current Executable Entry

Terminal A:
Terminal B:
If you want the current entry to use PostgreSQL-backed module wiring instead of in-memory stubs, start with Local PostgreSQL Development Setup.

​
Planned CLI Example (Not Yet Implemented)

​
Web UI Path

The resulting task state and audit result should later be visible through task and approval views in the Web UI.

​
Pitfalls / Risk Notes

  • an agent command is not a way to bypass authorization or approval
  • production commands should still be subject to the same step-up and approval guardrails