pyspur

Why we included this project

Anyone who has burned an afternoon chasing why an agent misbehaves will see the point of PySpur immediately. You assemble workflows on a visual graph, write the logic in Python or through the UI, and run each step while seeing exactly what it received and returned. The loop is built around test cases: define them up front, run them against your agent, and failures show up with the context that produced them instead of raw JSON in a terminal. It also handles the messy production details, from loops and human-in-the-loop checkpoints to multimodal inputs and trace history, so it holds up as a daily tool, not a demo. If you ship agents that need to be reliable and auditable, this turns guesswork into visible, repeatable steps.