langchain-visualizer

Why we included this project

When a LangChain agent or chain misbehaves, the hardest part is usually figuring out what actually happened inside the run. This library wraps your workflow and opens a browser UI that shows the full prompt text sent on every LLM call, color-codes which parts are hardcoded versus filled-in template variables, and lets you step through the execution flow to see when each function and tool is invoked. For multi-step agents, where one bad tool call or prompt substitution can derail the whole result, that kind of visibility is usually what's missing. It also reports the cost of each call and the total run when using OpenAI's text-davinci-003, which makes it easy to spot expensive or repeated calls while debugging. Teams already building on LangChain get a lightweight way to inspect and explain their pipelines without standing up a heavier tracing stack.