Coze Loop

Coze Loop helps developers test and monitor AI agents, with tools for investigating failed runs and comparing prompt changes. Its self-hosted edition brings prompt development, evaluation, and execution traces into one browser interface.

• Prompt testing and versions: Use a visual playground to compare outputs across LLMs, debug prompts, and keep track of prompt versions.
• Automated evaluation: Manage test datasets and evaluators, run experiments on prompts and Coze agents, and review results for criteria such as accuracy, conciseness, and compliance.
• Execution traces: Inspect model calls, prompt processing, and tool execution, including intermediate results and exceptions, using SDK-reported traces and traces captured automatically during playground sessions.

The free, open-source edition uses the Apache License 2.0 and runs on your own machine or server through Docker Compose or Kubernetes. A separate commercial edition is also available.

Model integrations use the Eino framework and include OpenAI and Volcengine Ark. Self-hosting the platform doesn't make those provider calls local: using these services requires their API credentials and sends model requests to the selected provider.