Linux / macOS

<p align="center">
  <img src="assets/hero-banner.png" alt="agenttrace — find where your AI agents waste money & time" width="100%">
</p>

<p align="center">
  <a href="https://github.com/luoyuctl/agenttrace/actions/workflows/ci.yml"><img src="https://github.com/luoyuctl/agenttrace/actions/workflows/ci.yml/badge.svg" alt="CI"></a>
  <a href="https://luoyuctl.github.io/agenttrace/"><img src="https://img.shields.io/badge/site-agenttrace-54ff00.svg" alt="Site"></a>
  <a href="https://github.com/luoyuctl/agenttrace/releases/latest"><img src="https://img.shields.io/github/v/release/luoyuctl/agenttrace?color=00ADD8" alt="Release"></a>
  <a href="https://pkg.go.dev/github.com/luoyuctl/agenttrace"><img src="https://pkg.go.dev/badge/github.com/luoyuctl/agenttrace.svg" alt="Go Reference"></a>
  <a href="https://goreportcard.com/report/github.com/luoyuctl/agenttrace"><img src="https://goreportcard.com/badge/github.com/luoyuctl/agenttrace" alt="Go Report Card"></a>
  <a href="https://github.com/luoyuctl/agenttrace/stargazers"><img src="https://img.shields.io/github/stars/luoyuctl/agenttrace?style=social" alt="GitHub stars"></a>
  <img src="https://img.shields.io/badge/go-1.24+-00ADD8.svg" alt="Go">
  <img src="https://img.shields.io/badge/license-MIT-green.svg" alt="License">
  <img src="https://img.shields.io/badge/Homebrew-v0.3.44-2bbc8a.svg" alt="Homebrew">
</p>

<h3 align="center">💸 Stop burning cash and hours on invisible AI agent waste</h3>

---

## What is agenttrace?

AI coding agents (Claude Code, Gemini CLI, Codex CLI) burn tokens in loops, retry failures silently, and leave you with a surprise bill. You're wasting **money** on dead tokens and **time** on broken sessions — and you can't even see where.

**agenttrace** finds the waste in both — so you stop paying for nothing and start shipping faster.

Site: https://luoyuctl.github.io/agenttrace/

Sample HTML report: https://luoyuctl.github.io/agenttrace/demo-report.html

Featured in: [Awesome Codex CLI](https://github.com/RoggeOhta/awesome-codex-cli), [Awesome Gemini CLI](https://github.com/Piebald-AI/awesome-gemini-cli), [Charm in the Wild](https://github.com/charm-and-friends/charm-in-the-wild), and [Awesome Claude Code and Skills](https://github.com/GetBindu/awesome-claude-code-and-skills).

<p align="center">
  <img src="assets/agenttrace-demo.gif" alt="agenttrace TUI demo" width="100%">
</p>

> Sharing agenttrace? See [docs/demo-playbook.md](docs/demo-playbook.md) for the GIF recording script, launch storyline, and posting copy.

## Why it exists

AI agents now behave like tiny build systems: they plan, call tools, retry, hang, and spend money while doing it. Most teams only see the final output, not the session health, token burn, tool failure rate, or whether the agent got stuck. agenttrace gives that missing operational view in the terminal.

## Where it fits

agenttrace is not a hosted tracing backend or another chat client. It is a local inspection tool for the logs your coding agents already leave on disk.

| If you need... | Use agenttrace for... |
|---|---|
| Local-first privacy | Inspect sessions without uploading prompts, code, or tool logs |
| Fast terminal triage | Open a TUI, sort bad sessions, and jump into detail/diagnostics |
| Cross-agent comparison | Compare Claude Code, Codex CLI, Gemini CLI, Qwen Code, Aider, Cursor exports, Hermes, OpenCode, Oh My Pi, Kimi, and more |
| Cost and token evidence | See cost, token usage, cache usage, retries, loops, latency, and health in one place |
| CI guardrails | Export JSON/Markdown/HTML and fail builds on low health or high tool failure rates |

## What it catches

| Pain | What agenttrace shows |
|---|---|
| Surprise token bills | input/output/cache tokens, estimated cost, top token-heavy agents |
| Silent tool loops | repeated tool calls, retry loops, long gaps, hanging sessions |
| Slow agents | P50/P95/P99 latency, per-tool latency ranking, timeout-like gaps |
| Quality regressions | health score, anomaly types, shallow reasoning, redacted thinking |
| Hard-to-compare tools | session diff across Claude Code, Codex CLI, Gemini CLI, Qwen Code, Aider, Cursor exports, Oh My Pi, and more |
| CI blind spots | JSON reports and health gates for average health, critical sessions, and tool failure rate |

## ✨ Features

| Feature | Description |
|---|---|
| 🚀 **Single Binary** | 7.5 MB — `curl -sL ... \| sh` install, no runtime deps |
| 🖥️ **Bubble Tea TUI** | Modern terminal UI: Overview → Session List → Detail → Diagnostics → Diff |
| ⚡ **Persistent Cache** | Incremental session cache avoids a full disk parse on every startup |
| 🩺 **Doctor Mode** | `--doctor` checks detected agent dirs, cache health, and next steps |
| ⌨️ **Command Mode** | `:health <80`, `:cost >0.1`, `:sort cost desc`, `:anomalies` |
| 🔍 **Multi-Format Auto-Detect** | Claude Code / Codex CLI / Gemini CLI / Qwen Code / Aider / Cursor exports / Hermes / OpenCode / OpenClaw / Oh My Pi / Kimi / Copilot-style logs |
| 💸 **Cost & 

... (truncated)