Who is operating
the agent?

A working description of the intended relationship between the ChatGPT-facing orchestrator, the local operational agent and a specialist project director.

Key propositionAn intended architecture
is not yet verified implementation.

The intended roles

Nickabot

The ChatGPT/Codex-facing interface intended to orchestrate work and communicate with the user.

Nitram

The OpenCLAW/OpenCore-side operational agent intended to manage local, longer-running work.

Project Director

A specialist Ends Up Here agent intended to be created and managed within that ecosystem.

The verification boundary

What has actually been created?

The architecture above describes the intended design discussed in the source conversation. It does not by itself prove that an agent exists on the desktop, that configuration files have been written, or that a runtime is connected.

A real confirmation needs concrete evidence: a file path, repository, configuration, runtime and a repeatable way to access the agent.

Working rule: treat design, conversation and implementation as separate states.

Why it matters

Agent identity is part of project accountability.

When an agent creates files, sends messages or changes a project, it should be possible to say which layer acted, under whose authority and where the durable result lives.

This note is a connected project record, not a claim that the intended system is fully operational. It should be updated when the architecture is verified against the local environment.

Questions to resolve

Where does Nickabot actually run?

What can Nitram change?

Where is the Project Director configured?

Which files are authoritative?

How is identity carried across agents?

What proves a task is complete?

Status note