Environmental Impact Tracker

A Claude Code skill that calculates and displays the environmental footprint of your AI interactions — translating abstract token usage into energy consumption (Wh), water usage (mL), and relatable real-world comparisons.

What it does

After heavy turns (5,000+ tokens, subagent use, or complex tasks), Claude automatically displays:

🌍 Environmental Impact of This Response

Energy: 22.1 Wh (about 1 phone charge)
Water:  37.6 mL (about 2.5 tablespoons)

Model: Claude Sonnet 5
Tokens: 25,000 in + 6,000 out + ~2,500 skill = 33,500
+ 136,000 agent tokens across 4 agents = 169,500 total

💡 If a lighter model had been used:
  Haiku: 7.6 Wh / 12.9 mL  (2.9x less)

Note: Estimates — see references/sources.md for methodology.

Features:

• Per-turn impact — energy and water for each heavy exchange
• Subagent tracking — includes all spawned agent activity in the total
• Skill cost included — accounts for the tokens used to load this skill itself
• Model efficiency comparison — shows what the same work would have cost on a lighter model
• Cumulative tracking — session, weekly, and project totals via a local log file
• Auto-wires enforcement — adds a rule to your CLAUDE.md on first use so it never gets skipped

Installation

Claude Code

/plugin install example-skills@anthropic-agent-skills

Or copy SKILL.md (and the references/ folder) into your Claude Code skills directory:

~/.claude/plugins/cache/anthropic-agent-skills/example-skills/unknown/skills/environmental-impact-tracker/

Manual (any Claude environment)

Paste the contents of SKILL.md into your system prompt or project instructions.

Usage

Once installed, the skill triggers automatically. You can also ask directly:

• "What was the environmental impact of that?"
• "Show me my session total"
• "Show me this week's impact"
• "Show me this project's total"

How it works

Energy and water estimates are calculated using published research on LLM inference compute costs:

Water usage: 1.7 mL per Wh (average data center cooling + power generation).

Sources: Epoch AI (2025), UC Riverside / Li et al. (2023). See references/sources.md for full citations.

Note: These are estimates derived from public benchmarks and pricing signals, not official Anthropic energy figures. Actual consumption varies by data center location, cooling method, and grid energy mix.

Keeping it up to date

Model rates are reviewed every 2 months. When new Claude models are released or new research is published, update the rates table in SKILL.md following the instructions in references/sources.md.

License

MIT