CrbonFree

Product Introduction

1. Definition: CrbonFree is a specialized AI carbon accounting and energy measurement platform. It is a SaaS (Software-as-a-Service) tool that provides granular, token-level metering of energy consumption and carbon dioxide equivalent (CO₂e) emissions for generative AI model inference.
2. Core Value Proposition: It exists to solve the critical problem of unmeasured and unreported environmental impact from AI usage. CrbonFree enables organizations to move from rough, spend-based estimates to auditable, per-call measurements, providing the data needed for accurate Scope 3 emissions reporting and sustainable AI governance.

Main Features

1. Granular, Per-Token Metering: CrbonFree's core technology meters every AI API call, disaggregating usage by model and provider. It counts input and output tokens, applying tiered, versioned emission factors (e.g., methodology v1.2) to convert token volume into energy (joules), carbon (CO₂e), and water usage. This provides a measurement with a defined ±28.3% uncertainty band, based on IPCC Tier 1 standards.
2. Multi-Provider & Multi-Access Integration: The platform connects to all major AI providers (OpenAI, Anthropic, Google Vertex AI, etc.) and gateways (OpenRouter, Vercel AI Gateway) via read-only API keys. It offers multiple integration surfaces: a TypeScript SDK (@crbonfree/sdk), Python SDK (crbonfree), a Model Context Protocol (MCP) server, a CLI for local agents, a browser extension for chat apps, and a REST API, ensuring tokens are counted wherever they occur.
3. Audit-Ready Reporting & Verification: CrbonFree generates defensible outputs for compliance. This includes signed digital retirement receipts linked to carbon credit serial numbers on registries like American Carbon Registry, publicly verifiable at verify.crbonlabs.com. It also produces monthly audit packs (ZIP files with verifiable manifest hashes) and provides badges/certificates for public claims, turning raw data into auditor-accepted evidence.

Problems Solved

1. Pain Point: The inability to accurately measure and report the carbon footprint of AI inference, which is a purchased cloud service falling under Scope 3 emissions. Current methods relying on cloud spend guesses are inadequate for tightening disclosure regulations and fail under auditor scrutiny.
2. Target Audience: Sustainability managers and ESG reporting teams needing accurate Scope 3 data; CTOs and engineering leaders implementing sustainable AI practices; developers and ML engineers building AI applications who require visibility into operational impact; and compliance officers preparing for mandatory carbon disclosure frameworks.
3. Use Cases: Essential for preparing annual sustainability reports (ESG); responding to detailed auditor questionnaires on IT and cloud emissions; conducting internal cost-vs-impact analysis of different AI models; and providing transparent environmental impact data to customers or stakeholders as part of corporate responsibility initiatives.

Unique Advantages

1. Differentiation: Unlike broad carbon accounting platforms or manual spreadsheets, CrbonFree is built specifically for the technical nuances of AI inference. It moves beyond high-level estimates by measuring actual token usage per call, unlike competitors that may only offer spend-based extrapolations or lack provider-specific granularity.
2. Key Innovation: Its methodology transparently separates and reports three distinct layers of impact: 1) the accelerator energy, 2) the data center facility overhead (via PUE), and 3) the embodied carbon from hardware and training. This layered, factor-versioned approach allows for precise comparison and eliminates the blending that causes disagreement in published AI emission figures.

Frequently Asked Questions (FAQ)

1. How does CrbonFree calculate AI carbon emissions? CrbonFree calculates emissions by first metering the exact number of input and output tokens used per AI model API call. It then applies published, versioned emission factors (e.g., 0.3 J/input token for medium-tier models) to calculate energy use. This energy is converted to CO₂e using grid carbon intensity and data center Power Usage Effectiveness (PUE), with additional layers for hardware and training carbon, following an open framework.
2. Is CrbonFree's data secure and private for my AI prompts? Yes. CrbonFree is designed for privacy. It only receives and processes usage metadata—specifically the model name, token counts, timestamps, and provider. The actual content of your prompts and completions never leaves your infrastructure or the provider's API; it is never sent to or stored by CrbonFree.
3. What AI providers and models does CrbonFree support for carbon accounting? CrbonFree supports direct integration with OpenAI, Anthropic, Google Vertex AI, AWS Bedrock, Microsoft Azure Foundry, and gateways including OpenRouter, Vercal AI Gateway, and Cloudflare AI Gateway. It uses a consistent metering methodology across all, with plans to add Cohere, xAI, Groq, Together AI, Replicate, and Ollama.
4. Can I use CrbonFree for free to measure my AI carbon footprint? Yes, CrbonFree offers a free plan that requires no credit card. It allows you to connect providers, meter token usage, and view associated energy, carbon, and water estimates on the dashboard. Features like signed retirement receipts, audit packs, and trust assets are available on paid tiers.
5. How does CrbonFree help with AI Scope 3 emissions reporting? Since AI inference runs on third-party infrastructure, it is a Category 1 "Purchased Goods and Services" Scope 3 emission. CrbonFree provides the specific, measured activity data (tokens) and applies defensible conversion factors to generate the CO₂e figure required for disclosure, replacing unverifiable spend-based estimates that auditors may reject.