# Binex

An open-source, local runtime for debugging AI agent workflows built as DAG pipelines, with tracing, replay and artifact lineage rather than a framework for building agents.

Binex is a local, open-source runtime for AI agent workflows built as directed acyclic graph pipelines. Rather than a framework for building agents, it focuses on what happens after a pipeline runs: tracing every step, tracking per-call cost, comparing runs, replaying and stepping back through a pipeline, and catching regressions against a saved golden run.

It ships a visual editor and a command-line interface, supports any model through LiteLLM including OpenAI, Anthropic, Google, Ollama, OpenRouter and DeepSeek, and can attach its observer to an existing LangChain, CrewAI or AutoGen pipeline to add the same debugging without migrating away from those tools. It installs from pip, with optional extras for LangChain, CrewAI, AutoGen, OpenTelemetry tracing and a colored CLI, and can also run in Docker.

Website: https://alexli18.github.io/binex