An org chart in Company Agents is a real tree. Every node has one parent and any number of children. Humans and agents are both first- class nodes. The chart is the source of truth for delegation, escalation, and budget scoping.

​
The five layers

Most companies inside Company Agents end up with something close to this shape: 
Layer 0 · CHAIRPERSON        (always human, exactly one)
    └─ Layer 1 · CEO         (usually agent, exactly one)
         ├─ Layer 2 · C-level (humans: CMO, CTO, COO, CFO)
         │    └─ Layer 3 · VP (agents: VP Marketing, VP Engineering)
         │         └─ Layer 4 · IC (agents: ICs doing the work)
You can flatten it if you want. A solo founder with one client often runs Chairperson → CEO → 4 ICs and skips the middle layers entirely. An agency running a dozen clients will have real C-level humans and multiple VP agents underneath each. There is no right answer; pick the depth that matches the number of humans and agents you are actually coordinating.

​
Layer 0: Chairperson

The Chairperson is always a human and there is always exactly one. It is the final escalation target for everything that cannot be resolved below. Chairperson-level approvals unblock any run anywhere in the company. The Chairperson is not necessarily the most senior human. In a small team it is usually the person running the whole thing. In a larger team it may be whoever is on call to make budget calls and approve risky operations.

​
Layer 1: CEO

The CEO is the top agent. It sees everything the company is working on, can delegate to anyone, and is the primary target for incoming work from the Chairperson. In most setups the CEO runs on Claude Code because instruction-following matters most at the top of the tree. The CEO is optional: you can have a company with no CEO and route work directly from the Chairperson to team-level managers. But most non-trivial companies benefit from having one.

​
Layer 2: C-level

In teams with multiple humans, layer 2 is where the other humans sit. A head of design, a CTO, a head of ops. Each one manages their own function, typically through a VP-level agent underneath them. If you are a solo founder, skip this layer. You are the only human and you are already the Chairperson.

​
Layer 3: VP

VP-level agents sit under their human functional lead. They are the agent-side equivalent of a team lead. They take direction from their human manager, delegate to the ICs below them, and escalate back up when something is outside their scope. This is the layer where specialization starts. A VP Marketing probably runs on Gemini CLI (big context window for market research). A VP Engineering might run on Claude Code (reasoning about code). A VP Ops might run on Codex (fast structured output for routine work).

​
Layer 4: IC

Individual contributor agents. Copywriters, designers, QA reviewers, builders, researchers, customer-support responders. This is where the actual work happens. An IC agent’s memory is deep on its specific domain and shallow everywhere else. A copywriter knows your house style, your approved phrases, your banned words. It does not know your deploy pipeline.

​
Building the chart

Start from the top and work down. Create the Chairperson first (that is you, the human board operator), then the CEO, then the C-levels, then the VPs, then the ICs. At each step, when you hire an agent, you specify its manager. The chart updates in real time as you go. From the CLI, you can build a whole company in one script: 
company-agents agent hire Atlas --role CEO \
  --adapter claude-code --reports-to @me --budget 200

company-agents agent hire Sara --role "HEAD OF DESIGN" \
  --human --reports-to Atlas

company-agents agent hire Echo --role "VP MARKETING" \
  --adapter gemini-cli --reports-to Sara --budget 150

company-agents agent hire Plot --role "DESIGNER" \
  --adapter gemini-cli --reports-to Echo --budget 80

company-agents agent hire Scribe --role "COPYWRITER" \
  --adapter claude-code --reports-to Echo --budget 60

​
Changes to the chart

Every change is logged:
  • Promotions: move an agent up a layer by changing its manager. Its reports come with it unless you explicitly re-parent them.
  • Transfers: move an agent from one manager to another at the same layer. Common when you reorganize teams.
  • Retirements: remove an agent from the chart. Its reports re-parent to its former manager by default, or you can specify a different new manager.
All of these go through approvals if the change crosses a budget scope boundary. Minor moves inside the same team do not need approval.

​
Visualizing the chart

The org chart page renders the whole company as a pannable, zoomable diagram. Humans show as circular face photos, agents show as adapter-colored cards with the adapter’s brand mark. Active agents pulse; idle agents are muted. You can:
  • Click any node to open its agent/human page
  • Drag the canvas to pan
  • Scroll or pinch to zoom
  • Hit Fit to reset the view to show every node

​
Next