OpenAI Usage and Costs

AITracer primarily tracks cost at the trace level.

For teams that need broader financial reporting, OpenAI also provides organization-level usage and cost APIs that can be used alongside AITracer trace records.

This allows teams to compare:

• trace-level execution costs
• organization-wide OpenAI spend
• model allocation trends
• project-level usage patterns
• cost anomalies across teams and workflows

AITracer can ingest this data and map it back to internal action groups, workflows, or business units.

Usage API

Use the OpenAI Usage API to retrieve model consumption data across your organization.

What it returns

• input tokens
• output tokens
• cached tokens
• audio tokens
• request volume
• project IDs
• user IDs
• model usage distribution
• API key attribution

Retrieval workflow

1. Create an OpenAI Admin API key
2. Request /v1/organization/usage/completions
3. Provide a start_time Unix timestamp
4. Use bucket_width such as 1d for daily aggregation
5. Use group_by=["model", "project_id"] when analyzing model distribution

import os
import time
import requests
 
headers = {
    "Authorization": f"Bearer {os.environ['OPENAI_ADMIN_KEY']}"
}
 
params = {
    "start_time": int(time.time()) - (30 * 24 * 60 * 60),
    "bucket_width": "1d",
    "group_by": ["model", "project_id"],
    "limit": 7,
}
 
response = requests.get(
    "https://api.openai.com/v1/organization/usage/completions",
    headers=headers,
    params=params,
)
 
response.raise_for_status()
 
usage_data = response.json()["data"]

Costs API

Use the OpenAI Costs API to retrieve organization-level billing totals.

What it returns

• total spend
• currency values
• line items
• project IDs
• billing buckets

Retrieval workflow

1. Request /v1/organization/costs
2. Use the same start_time window
3. Aggregate daily, weekly, or monthly cost buckets
4. Compare total spend against AITracer trace-level cost records

costs = requests.get(
    "https://api.openai.com/v1/organization/costs",
    headers=headers,
    params={
        "start_time": params["start_time"],
        "bucket_width": "1d",
        "limit": 30
    },
)
 
costs.raise_for_status()
 
cost_data = costs.json()["data"]

How AITracer uses this data

This data can be joined with AITracer records to answer broader operational questions:

• Which teams consume the most model resources?
• Which workflows generate the highest spend?
• Are trace-level costs aligned with provider billing totals?
• Are specific models driving abnormal spending patterns?

This helps organizations move beyond simple prompt-level cost tracking into broader financial governance.