aws-bedrock-cost-tool

aws-bedrock-cost-tool

A CLI tool for analyzing AWS Bedrock model costs using the Cost Explorer API.

Features

• Flexible time periods: days (7d), weeks (2w), months (1m), up to 365 days
• Multiple output formats: JSON, tables, ASCII plots, matplotlib plots (iTerm2), summary
• Three detail levels: basic, standard, full (with regional breakdown)
• Token usage and cache statistics
• AWS profile support
• Importable as Python library

Installation

Requirements: Python 3.14+, AWS CLI configured, uv

git clone https://github.com/dnvriend/aws-bedrock-cost-tool.git
cd aws-bedrock-cost-tool
uv sync
uv build
uv tool install dist/aws_bedrock_cost_tool-0.1.0-py3-none-any.whl

Configuration

AWS Credentials

# Option 1: AWS profile
aws configure --profile my-profile
aws-bedrock-cost-tool --profile my-profile

# Option 2: Environment variables
export AWS_PROFILE=my-profile

Required IAM Permission

{
  "Version": "2012-10-17",
  "Statement": [{
    "Effect": "Allow",
    "Action": "ce:GetCostAndUsage",
    "Resource": "*"
  }]
}

Cost Explorer must be enabled in AWS Console.

Usage

# JSON output (default), last 30 days
aws-bedrock-cost-tool

# Summary with all models, tokens, and cache stats
aws-bedrock-cost-tool --summary

# Table output
aws-bedrock-cost-tool --table

# Specific period and profile
aws-bedrock-cost-tool --period 7d --profile production

# Time series plot
aws-bedrock-cost-tool --plot-time

# Bar chart by model
aws-bedrock-cost-tool --plot-models

# All visualizations (ASCII)
aws-bedrock-cost-tool --all-visual

# High-quality matplotlib plots (iTerm2 only)
aws-bedrock-cost-tool --plot-image

# Full detail with regional breakdown
aws-bedrock-cost-tool --detail full --table

Output Examples

Summary:

Total: ~$1015.85

Models:
  Claude Sonnet 4.5: ~$962.35 (~24M tokens, cache read: ~1410M tokens, cache write: ~95M tokens)
  Claude Haiku 4.5: ~$35.92 (~24M tokens, cache read: ~20M tokens, cache write: ~5M tokens)
  Claude Sonnet 4: ~$16.95 (~4M tokens)
  Claude Opus 4.5: ~$0.63 (~0.006M tokens)

JSON (pipe to jq):

aws-bedrock-cost-tool | jq '.models[] | select(.model_name | contains("Sonnet"))'

Options

Option	Description
--period	Time period (7d, 2w, 1m, 3m). Default: 30d, Max: 365d
--profile	AWS profile name
--detail	basic, standard (default), full
--table	Table output
--plot-time	Time series plot (ASCII)
--plot-models	Bar chart by model (ASCII)
--all-visual	All ASCII visualizations
--plot-image	Matplotlib plots inline in iTerm2
--summary	Summary with tokens and cache stats
-v	Verbose logging (-vv for debug, -vvv for trace)
-q	Quiet mode

Library Usage

from aws_bedrock_cost_tool import (
    create_cost_explorer_client,
    query_bedrock_costs,
    analyze_cost_data,
    parse_period,
    calculate_date_range,
    format_date_for_aws,
)

days = parse_period("30d")
start_date, end_date = calculate_date_range(days)

client = create_cost_explorer_client(profile_name="my-profile")
response = query_bedrock_costs(client, format_date_for_aws(start_date), format_date_for_aws(end_date))
cost_data = analyze_cost_data(response, start_date, end_date, detail="standard")

print(f"Total: ${cost_data['total_cost']:.2f}")
for model in cost_data['models']:
    print(f"{model['model_name']}: ${model['total_cost']:.2f}")

Development

make install       # Install dependencies
make check         # Run lint + typecheck + test + security
make pipeline      # Full pipeline including build

License

MIT License - see LICENSE

Author

Dennis Vriend - @dnvriend