Cost Tracking¶
OneManCompany tracks LLM token usage and USD cost per project, so you always know what your AI company is spending.
What's Tracked¶
- Token usage — Input and output tokens per employee per task
- USD cost — Calculated from model pricing (varies by provider and model)
- Per-project breakdown — See exactly how much each project costs
How It Works¶
Every LLM call made by an employee is logged with:
- Employee ID
- Task/project context
- Model used
- Token count (prompt + completion)
- Estimated cost
Viewing Costs¶
Cost data is available in the CEO console. You can see:
- Per-task costs — How much each individual task consumed
- Per-project costs — Total cost across all subtasks in a project
- Per-employee costs — Which employees are most expensive
Cost Optimization Tips¶
Choose Models Wisely
Each employee can be assigned a different model. Use cheaper models for simple tasks (email, summaries) and more capable models for complex work (coding, design).
Use Claude Code for Heavy Work
If you have a Claude subscription, switching founding employees to Claude Code mode for complex tasks can reduce per-token OpenRouter costs.
Monitor Project Costs Early
Check costs during a project, not just after. If a task is burning through tokens, consider providing clearer instructions or breaking it into smaller pieces.
Cost vs Quality¶
Different models offer different cost/quality trade-offs:
| Tier | Example Models | Best For |
|---|---|---|
| Budget | Smaller open models | Simple tasks, high volume |
| Mid-range | Claude Haiku, GPT-4o-mini | Routine work, coordination |
| Premium | Claude Opus, GPT-4o | Complex reasoning, coding |
The flexibility to assign different models per employee lets you optimize this balance across your organization.