agentops.tools

Agent Trace Visualizer

Paste any agent trace — OpenAI tool-use, Anthropic Messages, LangSmith export, or OpenTelemetry GenAI spans. Get a clean timeline with token counts, costs per step, and retry loops auto-flagged. Everything runs in your browser.

Paste a trace and click Render, or load a sample.
Supports: OpenAI · Anthropic · LangSmith · OpenTelemetry GenAI

Supported formats

  • OpenAI — Chat Completions or Responses API JSON. Pass a single response or an array of multi-turn responses.
  • Anthropic — Messages API response with content blocks. Tool-use blocks become tool spans.
  • LangSmith — Trace export with run_type and child_runs. Token usage extracted from llm_output.
  • OpenTelemetry GenAI — Spans with gen_ai.* semconv attributes (gen_ai.request.model, gen_ai.usage.input_tokens, etc.).

What the colors mean

  • LLM call — model invocation with input/output tokens
  • Tool call — function/MCP invocation
  • Retry — same tool called 3+ times consecutively (loop heuristic)
  • Error — span returned non-OK status