Costs and budgets

You pay each provider directly. kunji helps you see what you’re spending and stop before it gets out of hand.

See costs as you go

Every answer ends with a line like:

claude-sonnet · 1,204 in · 380 out · $0.0093 · 3.4s

• /cost in a chat shows this chat’s cost, today’s cost and the model’s price.
• The bar at the bottom of the chat shows today’s total.

Usage reports

kunji usage          # last 7 days
kunji usage -d 30    # last 30 days

A table per provider and model with requests, tokens and cost. Usage is recorded on your computer only.

Daily budgets

kunji budget set openai 2      # OpenAI: max 2 per day
kunji budget set all 5         # all providers together: max 5 per day
kunji budget show
kunji budget clear openai

• At 80% of a budget, kunji warns you once per chat.
• At 100%, kunji stops sending messages to that provider until tomorrow (or until you raise the limit).
• Amounts are in your display currency.

Caution

kunji’s numbers are estimates based on public price lists. Your provider’s billing page is always the real number. Many providers also let you set a hard spending limit in their dashboard; it’s a good idea to set one there too.

Show costs in your currency

kunji config set display.currency INR
kunji config set display.usd_rate 84      # how many rupees per 1 US dollar

Or Settings → Currency in the menu. Prices are tracked in US dollars and converted with your rate, so update it now and then.

When a price is unknown

If kunji doesn’t know a model’s price, the cost shows as — and the usage table marks it with +?. You can set the price yourself (US dollars per million tokens) in config.toml:

# Example only: copy the real numbers from the provider's pricing page.
[pricing."groq:llama-3.3-70b-versatile"]
input = 0.59    # US dollars per 1 million input tokens
output = 0.79   # US dollars per 1 million output tokens

Take the numbers from the provider’s pricing page. See How costs are calculated.