Koskamo
Navigation

Estimate AI model usage costs

Get an accurate cost estimate for running specific AI tasks on different models, accounting for tokenization differences and input/output splits, helping to budget for AI usage.

How it works

Read-onlyFinance and payments

The agent uses the estimate tool (via MCP or REST) to calculate costs for a given text input and expected output tokens. It passes the text and token counts to the server, which handles the tokenization and multiplication by current prices. The agent reviews the costSplit to understand if costs are driven by input or output and warns that cache/batch discounts are not applied.

Prompt

Paste this into an agent that has the servers below connected. Adjust the names and dates to your own.

Prompt for Estimate AI model usage costs
Estimate the cost of processing a 1000-word document with an expected output of 500 tokens using 'openai/gpt-5.2' and 'anthropic/claude-opus-4.6'. Use the AI Wave MCP server for the calculation.

MCP servers it needs

Source

The vendor page this playbook is drawn from, and the passage that supports it.

Cost estimate Works out what your usage costs on each model. You could do the multiplication yourself from the price table, but counting tokens is the part that does not travel
API and MCP | AI Wave

Each playbook is drawn from the vendor page it cites and reviewed by Koskamo before it is published. Koskamo does not run these servers or your agent: check what each tool will do before you let it act on your accounts.

More playbooks like this