Your Configuration
Number of AI agents
10
Separate agents or services making LLM calls
Calls per agent per day
100
Average LLM API calls per agent daily
Avg. tokens per call
2,000
Total input + output tokens (70/30 split assumed)
Monthly Cost Estimate
Monthly Total
all agents combined
Cost per Agent
per month
Daily Total
across all agents
Cost per 1K Calls
at current model
▼ Routing savings potential
If you route 70% of calls to a cheaper model, you could save this amount monthly.
total API calls/month · tokens/month
Model cost comparison — same workload
Cost calculated at your current inputs. Pricing approximate, updated April 2026.

Stop estimating. Start tracking actual spend.

This calculator uses averages. Costline shows you real per-agent costs pulled directly from OpenAI and Anthropic — by agent, by model, by day.

Start tracking free →
Free for 2 providers · No credit card required

Get this calculator as a shareable report

Enter your email and we'll send you a formatted cost estimate you can share with your team.

How this calculator works

Token split: We assume 70% input tokens and 30% output tokens per call — a reasonable average for most agent workloads. Adjust as needed for your use case (output-heavy generation tasks will be more expensive).

Pricing: Based on published April 2026 list prices: GPT-4o at $2.50/$10 per 1M input/output, GPT-4o mini at $0.15/$0.60, Claude 3.5 Sonnet at $3.00/$15.00, Claude Haiku at $0.25/$1.25, Gemini Pro at $1.25/$5.00. Prices vary with volume discounts and change over time.

Savings estimate: The routing projection assumes 70% of calls are routed to the cheapest available model in the same family (or the cheapest model overall), with the remaining 30% staying on your primary model. Real savings depend on task mix and quality requirements.

Want actual numbers? Connect your API keys in Costline — we pull real usage data directly from OpenAI and Anthropic and show per-agent breakdowns.