LLM Usage Expenses
Keito includes a built-in “LLM Usage” expense category for tracking the cost of AI model usage. This category is automatically created in every workspace.
How It Works
LLM usage expenses use the unit-based expense model:
quantity (tokens in thousands) × unit_price = total_costFor example: 45k tokens at £0.003/k = £0.135.
The expense flows through the standard pipeline: it’s allocated to a project, included in reports, and can be invoiced to clients.
When LLM Expenses Are Created
- Automatically — when an agent integration logs token usage via the API or SDKs.
- Manually — you can create LLM usage expenses in the web UI like any other expense.
The current Keito CLI tracks time only. Use the API, Node SDK, Python SDK, or web app for LLM usage expenses.
What’s Captured
| Field | Description |
|---|---|
project_id | The project the tokens were used for |
quantity | Token count (typically in thousands) |
unit_price | Cost per unit |
total_cost | Calculated total |
notes | Model name, context (e.g., “claude-opus-4-6: 25k input + 15k output”) |
metadata | Structured token breakdown, session ID |
Viewing LLM Costs
- Go to Expenses and filter by category → “LLM Usage”.
- Use the source filter to see only agent-created expenses.
- In Reports → Expense Reports, group by category to see total LLM spend across projects.
Invoicing LLM Costs
LLM expenses can be included on client invoices as a separate line item or grouped with other project expenses. See Invoicing Agent Work for details.