// agent_trace_visualiser

Paste a Claude or OpenAI agent session JSON and get an interactive timeline of every message, tool call, tool result, and token count — step by step.

// trace_json_input

— model

— messages

— total tokens

— tool calls

about_this_demo.sh

$ cat formats.md

Accepts: Anthropic Messages API (input_tokens/output_tokens), OpenAI Chat

Completion (prompt_tokens/completion_tokens), raw messages arrays, Claude

extended thinking blocks, tool_use / tool_result pairs, and OpenAI tool_calls.

$ cat use_cases.md

Debug why a multi-step agent looped or hallucinated a tool call. Audit token

spend per step. Review the reasoning chain between a user prompt and the final

response. Verify tool inputs and outputs in a clinical or agentic workflow.