Cost

The Cost page shows your LLM spend broken down by agent and time period.

Cost chart

A daily bar chart shows total spend over the last 30 days. Hover over any bar to see the exact amount for that day.

Cost by agent

Below the chart, each agent shows its cumulative cost for the selected period. Use this to identify which agents are the most expensive to run.

How cost is calculated

Cost is computed from the token counts and model pricing reported by your LLM provider. For each LLM call:

cost = (prompt_tokens / 1000 * input_price) + (completion_tokens / 1000 * output_price)

Costs are aggregated at the session level and then at the agent level.

Cost anomaly alerts

If a single session costs more than $1.00, Maev classifies it as a Cost anomaly with high severity. Sessions over $5.00 are classified as critical.

You will receive an alert immediately when a cost anomaly is detected.

Reducing cost

Common causes of unexpected cost:

• Infinite loops: the agent keeps calling the same LLM API repeatedly
• Oversized prompts: the system prompt or context window is growing unboundedly
• Wrong model selection: using a large model for a task that a smaller one can handle

Check the Patterns page to see if cost anomalies are recurring.