# SOMA AI Emission Factors (SAEF)

> SOMA AI Emission Factors (SAEF) is an open-source MCP server for climate and sustainability. Open, versioned carbon and water factors for AI inference, for Scope 3 reporting. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** API, CLI, self-hosted, Web
- **Website:** https://somaai.earth/ai-footprint
- **Repository:** https://github.com/guillermollopis/soma-ai-emission-factors
- **MCP Registry name:** io.github.guillermollopis/saef
- **Categories:** climate and sustainability
- **Listing updated:** 2026-11-18

## About

The README says SAEF is a dataset of estimated carbon and water factors for AI inference, broken down by model size class and by cloud region or national grid. Its figures are built to feed Scope 3 Category 1 accounting under the GHG Protocol and CSRD, and every factor is traced back to a file and a cited source.

The factors are applied to token counts, so a team can convert an API invoice or usage export into a kg CO2e estimate with a central value and a high case. Water is reported alongside carbon for 17 regions and is never added to the carbon total. The project also ships a Python package called saef, a command-line tool and an MCP server, plus LiteLLM and LangChain integrations that estimate each call as it happens.

The listing is aimed at sustainability and finance teams preparing emissions inventories, and at developers who want per-call estimates in their own code. The documentation stresses that all results are estimates, not meter readings, and that provider energy use for hosted models cannot be measured directly by customers.

## Features

- **Model class factors:** Factors for small, mid-size and frontier model classes, with a mapping of common model identifiers to each class.
- **Regional grid factors:** Carbon intensity by cloud region, country and US eGRID subregion, covering 210 countries and economies.
- **Token-based calculation:** Converts billed or effective token counts into kg CO2e, with a cache credit and a conservative high case.
- **Water estimates:** On-site and off-site water factors for 17 regions, reported beside carbon rather than added to it.
- **Python package and CLI:** A dependency-free saef package and command-line tool for running estimates in scripts or terminals.
- **LLM framework integrations:** LiteLLM, LangChain callback and middleware hooks that estimate emissions per model call.
- **MCP server:** Exposes estimate, chat usage, region comparison and model class tools to AI clients over stdio.
- **Versioned, citable dataset:** Released under CC BY 4.0 with a version DOI and concept DOI so past inventories can be reproduced.

## Good for

- Estimating Scope 3 Category 1 emissions from a company's purchased AI services
- Logging per-call carbon and water estimates inside an application that calls LLM APIs
- Comparing the footprint of the same token workload across cloud regions
- Adding an emissions estimate to a team's ChatGPT, Copilot or Claude seat usage
- Documenting factor versions and assumptions for a sustainability or CSRD report

## Install

### Claude Code (Terminal)

```sh
claude mcp add saef -- uvx saef
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "saef": {
      "command": "uvx",
      "args": [
        "saef"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "saef": {
      "command": "uvx",
      "args": [
        "saef"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "saef": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "saef"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "saef": {
      "command": "uvx",
      "args": [
        "saef"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "saef": {
      "command": "uvx",
      "args": [
        "saef"
      ]
    }
  }
}
```

## Tools

- `estimate_footprint`: Returns carbon, energy and water estimates for a token count, model and region.
- `estimate_chat_usage`: Returns the same estimates for chat exchanges using the dataset's default token size.
- `model_class`: Returns the size class letter and description for a model identifier.
- `list_regions`: Returns accepted region codes with names for cloud regions and countries.
- `compare_regions`: Returns carbon and water per region for a token count, sorted from lowest to highest carbon.

## FAQ

### Can the energy use of a hosted AI provider be measured directly by a customer?

No. The README says providers do not publish per-customer energy for requests, so customers can only estimate it from usage, model size and grid.

### How do I install the saef package?

Run pip install saef. The README says the core package has no dependencies; the MCP server needs the optional mcp extra.

### What license applies to the data and the code?

The data is licensed under CC BY 4.0 and the Python package code under MIT, according to the README.

### Which pricing applies?

The README does not describe any paid plan for the dataset or package, and the free SOMA AI calculator is mentioned as a separate tool.

### Is SOMA AI Emission Factors (SAEF) free?

Yes. SOMA AI Emission Factors (SAEF) is open source, so you can use and self-host it for free.

### How do I add SOMA AI Emission Factors (SAEF) to Claude Code?

Run this command in your terminal: claude mcp add saef -- uvx saef

### What are the best alternatives to SOMA AI Emission Factors (SAEF)?

The closest alternatives to SOMA AI Emission Factors (SAEF) are Bloomy Earth and Electricity Maps.

## Where this comes from

- Description: written from the project's README, the project's website and its MCP Registry entry.
- Install steps: from the official MCP Registry entry.

## Alternatives

- [Bloomy Earth](https://ai-powered.com/mcp/bloomy-earth): Bloomy Earth is a free MCP server for climate and sustainability.
- [Electricity Maps](https://ai-powered.com/mcp/electricity-maps): Electricity Maps is a freemium MCP server for maps and travel.

---
Source: https://ai-powered.com/mcp/soma-ai-emission-factors-saef
