Major Release

The Long-Term Supported version of Umbraco is here

Umbraco 17 LTS release →

Introducing the Umbraco Sustainability Agent Profile

A practical starting point for lower-carbon development and lower costs

Martin Wülser
Written by Martin Wülser

Large language models are already an established part of daily development workflows. They help with research, coding, troubleshooting, and documentation. At the same time, every generated token carries a real environmental and monetary cost. As it is tradition in the Umbraco community we are working with open collaboration and responsible engineering. AI assisted development is here to stay but we have an opportunity to shape how these tools are used in Umbraco projects.

Thomas Morris from the Umbraco Community Sustainability team has prepared a first draft of an “agents.md” file. It is a lightweight set of instructions designed to help developers use AI tools in a more efficient way that results in lower-emissions, while still getting reliable help with Umbraco development.

The intention is that not only will it conserve tokens as you use your LLM. Ultimately the idea is that it should also reduce running costs of the solutions you are creating with it as more efficient, less carbon-emitting solutions frequently also come with lower resource use and thus lower running costs.

This file is now available for community use, and we encourage everyone working with Umbraco to try it out.

Why create an AI agent profile for Umbraco?

The goals behind the draft are simple:

  • Provide an LLM with specific guidance so it behaves like a focused Umbraco assistant

  • Put sustainability at the center of AI-supported development

  • Reduce unnecessary token usage

  • Promote development practices aligned with the Web Sustainability Guidelines and the Umbraco Sustainability team’s recommendations

  • Offer a shared baseline that we can iterate and improve together

This approach helps ensure that when we rely on AI tools, we do so in a way that is as sustainable as possible .

What the agent does

When used in tools like ChatGPT or GitHub Copilot, the agent.md profile guides the model to:

1. Focus on Umbraco documentation

The agent knows to base its help on the official Umbraco documentation and match responses to the specific CMS version used in the project.

2. Apply sustainable web development practices

It actively encourages choices that reduce energy and resource consumption, such as:

  • Optimising images and media

  • Reducing HTTP requests

  • Keeping page weight under control

  • Lazy loading assets

  • Avoiding unnecessary dependencies

  • Using modern image formats

  • Monitoring Core Web Vitals

  • Reducing or removing unused content and media

3. Keep responses lean

The profile limits unnecessary verbosity and focuses on immediate, actionable guidance. This reduces token consumption and improves clarity.

How you can use it

You can drop the agents.md file directly into your project or personal AI setup. Most tools allow setting a system prompt or project-level instruction file. Once active, the model will automatically follow the sustainability-focused Umbraco behaviour outlined in the file.

Teams can also adapt the file to match their own coding standards or internal review processes.

This is a starting point. We need your input.

The draft is intentionally lightweight. It is meant to grow with feedback from developers using it in everyday work.

We would love to hear from you if you:

  • Notice gaps

  • Have suggestions for clearer guidelines

  • Find a sustainability practice that should be added

  • Or have practical examples from your projects

The Umbraco Sustainability Community team will continue developing this resource based on your experiences.

Moving Umbraco development toward lower carbon-emissions

AI-assisted development is here to stay. By shaping the tools rather than letting them shape us, we can keep Umbraco development productive while reducing the environmental impact of our workflows. This agent profile is one step in that direction.

We invite you to test the agents.md file, and share your thoughts. Together we can evolve a practical guide that supports high-quality, sustainable Umbraco solutions for years to come.