LLM Agent Trace Visualizer

Trace Visualizer parses and displays LLM agent execution traces in a readable timeline format. Multi-agent systems generate complex logs with tool calls, reasoning steps, messages between agents, and nested sub-tasks. Raw JSON traces are hard to read — this tool normalizes them into a color-coded timeline that shows the flow of messages, tool invocations, and decisions. It handles LangChain LCEL event streams, AutoGen message arrays, and any JSON array with recognizable fields (role, type, content, tool, action, agent).

The visualizer parses the input JSON array and classifies each entry by its structure: system messages, user messages, assistant responses, tool calls, tool results. Entries are rendered as a chronological list with color coding by type. Tool calls show the function name and arguments; tool results show the return value. Nesting is preserved for parallel or nested sub-agents.