We just open-sourced AgentTrace, a lightweight tracing and evaluation framework for AI agents and LLM workflows. Originally built as an internal tool at TensorStax, it helps you trace function calls, monitor OpenAI API usage, tag sessions, and evaluate model outputs with custom scoring logic.
It runs locally by default (SQLite), comes with a built-in web dashboard, and is easy to plug into async + tool-using workflows.
If you’re building agents or working with LLMs in production, we’d love feedback and contributions.
It runs locally by default (SQLite), comes with a built-in web dashboard, and is easy to plug into async + tool-using workflows.
If you’re building agents or working with LLMs in production, we’d love feedback and contributions.
GitHub: https://github.com/tensorstax/agenttrace by ariaattar 1 hour ago