Monitor your cloud's carbon footprint in real-time
Cloud Carbon Exporter automatically discovers cloud resources, estimates energy consumption and carbon emissions in real-time. This tool provides valuable insights for operational and tech teams interested in following the Carbon-Driven Development principles.
Carbon-Driven Development (CDD) is a philosophy to build digital services and manage cloud infrastructure with environmental sustainability at its core.
It revolves around three pillars:
- Estimate energy consumption for each cloud resource (servers, load balancers, storage, etc.)
- Collect data in production environments
- Aggregate data in real-time
By applying these few rules, production teams will be able to:
- Measure the overall energy efficiency of a system in relation to a business use (active user, transaction, etc.).
- Detect infrastructure anomalies faster
- Engage the company's operational teams and Execs more widely in continuous improvement
- Reduce the carbon footprint of applications
Check out the original article which explains in detail the concepts of CDD.
On the screenshot above, you can visualize and understand easily:
- the estimated energy consumed by connected user to the online service,
- the current CO2 emissions,
- the equivalent in turned on lightbulbs.
You can easily customize the content of this dashboard using the data returned by the exporter.
Try our live demo with our Grafana dashboard: https://demo.carbondriven.dev
Multi Cloud · We want to support as much cloud platform as possible. From hyperscalers to edge datacenters to regional provider. For now we support: GCP, AWS, Scaleway
Multi Model · The exporter is designed to be model agnostic. We plan to support and contribute back to:
- https://www.cloudcarbonfootprint.org/
- https://boavizta.org/en
- Other experimentations
This tool will prioritize the number of supported resources over the precision of the exported metrics. Estimating precisely the energy consumption of a resource is a hard task. The complexity and opacity of a Cloud service increase the margin of error but trends should be respected.
Once the resource energy draw is estimated, the exporter evaluates the carbon intensity of the resource at its location based on publicly available datasets.
OpenMetrics · The exporter is compatible OpenMetrics format. Therefore, you can ingest metrics into Prometheus, Datadog and every time series database that support this standard.
Performance · We're paying close attention to the exporter performance. Most API requests are done concurrently and cached. Most scrapes finish under 1000ms even with thousand monitored resources.
You can download the official Docker Image on the Github Package Registry
$ docker pull ghcr.io/superdango/cloud-carbon-exporter:latest
The Cloud Carbon Exporter can work on Google Cloud Platform, Amazon Web Service and Scaleway (more to come).
sequenceDiagram
Prometheus->>cloud-carbon-exporter: scrape metrics
cloud-carbon-exporter->>Asset Inventory: Query all used resources
cloud-carbon-exporter->>Monitoring: Get Resource statistics
cloud-carbon-exporter-->>Prometheus: Returns Watts and CO2 metrics
The exporter uses GCP Application Default Credentials:
GOOGLE_APPLICATION_CREDENTIALSenvironment variablegcloud auth application-defaultlogin command- The attached service account, returned by the metadata server (inside GCP environment)
$ docker run -p 2922 ghcr.io/superdango/cloud-carbon-exporter:latest \
-cloud.provider=gcp \
-gcp.projectid=myproject
sequenceDiagram
Prometheus->>cloud-carbon-exporter: scrape metrics
cloud-carbon-exporter->>AWS Cost Explorer API: Query used services and regions
cloud-carbon-exporter->>AWS Resources API: Describe Resource
cloud-carbon-exporter->>Cloudwatch API: Get Resource statistics
cloud-carbon-exporter-->>Prometheus: Returns Watts and CO2 metrics
The exporter is:
- Environment Variables (
AWS_SECRET_ACCESS_KEY,AWS_ACCESS_KEY_ID,AWS_SESSION_TOKEN) - Shared Configuration
- Shared Credentials files.
$ docker run -p 2922 ghcr.io/superdango/cloud-carbon-exporter:latest \
-cloud.provider=aws
sequenceDiagram
Prometheus->>cloud-carbon-exporter: scrape metrics
cloud-carbon-exporter->>Regional APIs: Query all used resources
cloud-carbon-exporter->>Cockpit: Get Resource statistics
cloud-carbon-exporter-->>Prometheus: Returns Watts and CO2 metrics
Configure the exporter via:
- Environment Variables (
SCW_ACCESS_KEY,SCW_SECRET_KEY)
$ docker run -p 2922 ghcr.io/superdango/cloud-carbon-exporter:latest \
-cloud.provider=scw
Cloud Carbon Exporter can easily run on serverless platform like GCP Cloud Run or AWS Lambda for testing purpose. However, we do recommend running the exporter as a long lived process to keep its cache in memory (lowering the cost)
Usage of ./cloud-carbon-exporter:
-cloud.aws.defaultregion string
aws default region (default "us-east-1")
-cloud.aws.rolearn string
aws role arn to assume
-cloud.gcp.projectid string
gcp project to explore resources from
-cloud.provider string
cloud provider type (gcp, aws, scw)
-demo.enabled string
return fictive demo data (default "false")
-listen string
addr to listen to (default "0.0.0.0:2922")
-log.format string
log format (text, json) (default "text")
-log.level string
log severity (debug, info, warn, error) (default "info")
Environment Variables:
SCW_ACCESS_KEY
scaleway access key
SCW_SECRET_KEY
scaleway secret key
Calls to cloud monitoring APIs can incur additional costs. The exporter will do its best to cache API responses and therefore, lower the impact on your bill. API costs are directly correlated to the number of resources the exporter generate data from. Here are the average costs you may observe per resource on your cloud account or project (instance, bucket, load balancer) for a 15 minutes cache TTL:
- AWS: $0,06 / month per resource
- GCP: $0,03 / month per resource (will be 10 times less in October 2025)
- SCW: free
The prices shown below are dated March 2025 and are subject to change by the cloud providers.
You can use the Cost Calculator file to do finer estimations with your own inputs. In this file, you can also anticipate the storage cost of carbon metrics if you choose to use the cloud provider monitoring service.
The exporter requires permissions to automatically discover resources in your cloud environment. For a quick and easy setup, you can grant it read-only access to your entire cloud platform, such as the Project Viewer role or ViewOnlyAccess policy.
If you'd prefer a more precise approach, you can authorize only the specific API calls needed for the services you use. A detailed list of required permissions for each cloud provider service will be available soon.
If the exporter encounters a missing permission, it will log a warning with details about the issue and increment the error_count{action="collect"} value. We recommend periodically monitoring this metric and adjusting permissions as needed to ensure smooth operation.
go build \
-o exporter \
github.com/superdango/cloud-carbon-exporter/cmd && \
./exporter -cloud.provider=aws -log.level=debug
We're grateful for every contribution that helps shape Cloud Carbon Exporter. Whether it's through testing, feedback, or documentation, each effort strengthens our software and enhances the user experience.
We'd like to extend our heartfelt appreciation to the individuals who have invested significant time and energy into making this project better.
- Thanks to Hakim Rouatbi, Raphaël Cosperec, Souhail Hanfi, Jonathan Chardon for giving early feedback.
We appreciate your input and contributions to Cloud Carbon Exporter. Here's how you can help:
Share Feedback and Ideas · Found a bug or have a feature idea? Start a discussion in our GitHub Discussions. Your testing and feedback are crucial to improving the project.
Code Contributions · We're actively refactoring to improve the codebase. For now, we're focusing on smaller, targeted contributions to ensure a smooth integration.
Model Contributions · Contribute directly to source model like Boavizta or Cloud Carbon Footprint
Thank you for your support!
dangofish.com - Tools and Services for Carbon-Driven Developers.
This software is provided as is, without waranty under AGPL 3.0 licence