How it works

From “Go” to Approved, Every Step Is Logged.

Here is what happens when an agent picks up work—whether you asked, a schedule fired, or another agent handed it off.

1

Work starts four ways

You ask in chat or by voice. A schedule fires—the daily executive briefing, the weekly spend review. An integration event arrives. Or another agent asks for help: the CMO asks the Data Analyst for numbers, the Social Media Manager asks the Graphic Designer for a visual.

Screenshot to come

Agent chat with mic, plus scheduled workflows

Screenshot to come

Agent detail — role, goals, tools, model policy

2

The agent plans, then checks its permissions

Each agent carries a list of tools it may use and tools it may not. The LLM Manager picks the model for the task from a registry—by cost, capability, and how sensitive the data is. Strict mode keeps private material off third-party models.

3

It pulls memory and your training

Business goals, past decisions, your style, examples you marked strong or weak, and what happened the last time it tried something similar. Every source that shaped the output is cited so you can see why it did what it did.

Screenshot to come

Memory view with citations

Screenshot to come

Approval queue

4

Risky actions stop at the approval queue

Spending, publishing, sending email, deleting, changing credentials or permissions—each arrives as a request with the proposed action, expected outcome, cost, and supporting evidence. You approve, deny, or send it back for revision.

Everything else—analysis, recommendations, drafts, internal tasks—just gets done.

5

It’s logged, and the agent learns

Every action, agent-to-agent message, tool call, and model call lands in the audit log with its cost. Your approval, correction, or rating goes into that agent’s training memory—so the next time is closer to how you would have done it.

Screenshot to come

Audit log and training panel

Agents do this on their own

Read analytics, social, finance, and website data

Write reports and daily briefings

Draft posts, scripts, emails, and invoices

Create and update tasks and notes

Ask other agents for help

Agents wait for you

Spend money

Publish anywhere

Send external email or start a campaign

Delete anything

Change credentials, permissions, or integrations