Home
New to AI
All 47 tools →
Features
Use cases
Live demo
Impact analytics
Pricing
Blog
FAQ
About
CLAUDE.md Generator
Few-Shot Example Set Builder
MCP Config Wizard
System Prompt Generator
Prompt Optimizer
RAG Chunking & Retrieval Planner
ReAct Prompt Converter
Skill File Generator
Structured Output Schema Builder
Subagent & Slash Command Generator
Tool / Function Schema Generator
Workflow to Agent Blueprint
Free tool
Agent Trace Analyzer
Paste an agent trace and find the loops, retries, and discarded work that quietly inflate your AI bill.
Get started-
Finds the five most common causes of surprise agent bills automatically — duplicate/near-identical calls, retry loops, discarded tool results, unbounded context growth, and steps that never fed into the final answer.
-
Every finding names the exact steps responsible and a dollar figure, never a vague "this looks inefficient".
-
Projects the waste at your real session volume, so a per-session finding becomes a monthly number worth acting on.
Frequently asked questions
The analysis itself (parsing, per-step cost, and every waste detection) is deterministic PHP — no model call. The optional "Write a prioritised fix list" button uses Claude (Anthropic) to turn those already-computed findings into a short narrative.
A documented canonical JSON shape (a list of steps with role/type/tool/tokens/model), an Anthropic Messages API-style transcript, and an OpenAI chat-completions-style log are auto-detected. Anything else is mapped heuristically from common key names — see the supported-formats panel on the run page.
Yes. Agent Trace Analyzer is a Pro-plan tool.
The parsing, pricing, and every waste detection run entirely on our server without ever calling a model. Only if you click "Write a prioritised fix list" is a summary of the already-computed findings (numbers and short descriptions, never your raw trace content) sent to Claude.