Cozeloop
Coze Loop is an open-source, developer-oriented platform for the full lifecycle management of AI agents, covering development, debugging, evaluation, and monitoring. It solves the challenges of prompt engineering, agent evaluation, and post-deployment observability, and is designed for developers building and operating AI agents.
✨ Key features
- Prompt development with visual Playground and version management.
- Systematic evaluation with automated multi-dimensional testing.
- Full execution observability from input to output.
- Integration with OpenAI, Volcengine Ark, and other models.
- Open-source edition with customization and extension capabilities.
🎯 Use cases
- Develop and debug prompts interactively with different LLMs.
- Evaluate prompt and agent outputs for accuracy, conciseness, compliance.
- Monitor and trace agent executions in production.
- Customize and extend the platform for specific business needs.
📦 Installation
🧰 Requirements: Requires Docker Engine for Docker Compose deployment, or Kubernetes cluster with Nginx Ingress for Helm deployment. Model API keys (e.g., Volcengine Ark) are needed for model configuration.
Docker Compose
- Clone the source code:
git clone https://github.com/coze-dev/coze-loop.git
cd coze-loop
- Configure a model: edit
release/deployment/docker-compose/conf/model_config.yamland setapi_keyandmodelfields. - Start the service:
make compose-up
- Access via browser at
http://localhost:8082.
Kubernetes (Helm)
- Pull the Helm chart:
helm pull oci://docker.io/cozedev/coze-loop --version 1.0.0-helm
tar -zxvf coze-loop-1.0.0-helm.tgz && cd coze-loop && rm -f ../coze-loop-1.0.0-helm.tgz
- Configure model in
release/deployment/helm-chart/umbrella/conf/model_config.yaml. - Configure Ingress rules in
templates/ingress.yaml. - Deploy:
make helm-up
make helm-pod
make helm-logf-app
make helm-logf-nginx
- Access via assigned domain.
🚀 Usage
After deployment, access the Coze Loop UI via browser (e.g., http://localhost:8082 for Docker). Then use the platform to create and debug prompts, set up evaluations, and view traces. For SDK usage, refer to the examples in the examples/ directory and modify values.yaml for Helm deployments.
⚠️ Good to know
Public network deployment has security risks including account registration, SSRF, and horizontal privilege escalation; assess and protect accordingly.
❓ FAQ
What are the deployment options?
Coze Loop can be deployed using Docker Compose or Kubernetes with Helm Chart.
Which models are supported?
It supports integration with OpenAI, Volcengine Ark, and other models via the Eino framework.
Is there an open-source edition?
Yes, Coze Loop offers an open-source edition with core features free to use and customizable.
How can I contribute?
Contributions are welcome via GitHub Issues, Pull Requests, and community discussion groups.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.