agenttrace
agenttrace dashboard showing AI agent observability metrics

Find AI agent waste before it ships.

agenttrace turns Claude Code, Codex CLI, Gemini CLI, Qwen Code, Aider, Cursor exports, Hermes, OpenCode, Kimi, and Copilot-style logs into one fast local TUI. 

curl -sL https://raw.githubusercontent.com/luoyuctl/agenttrace/master/install.sh | sh
go install github.com/luoyuctl/agenttrace/cmd/agenttrace@latest
agenttrace --demo
agenttrace --demo --overview -f html -o report.html
Token bills Input, output, cache tokens, estimated cost, and top burning sessions.
Tool loops Repeated calls, retries, long gaps, hanging runs, and tool failures.
Regressions Health scores, shallow reasoning, redactions, anomalies, and diffs.
CI evidence JSON, Markdown, and HTML reports with health and tool failure gates.

A terminal workbench for agent sessions.

Sort hot sessions, drill into details, compare nearby runs, and inspect diagnostics without uploading logs to another service.

agenttrace TUI session list with health filtering and cost metrics
0 Overview

See the operating picture.

Sessions, tokens, cost, error rate, P95 latency, health score, recent anomalies, and top agents.

1 Session List

Find the expensive run.

Sort by health, cost, or turns; filter by source, model, anomaly, text, or health threshold.

2 Detail

Explain what went wrong.

Primary issue, impact, evidence, next action, loop analysis, tool warnings, and cost alerts.

3 Diag / 4 Diff

Turn symptoms into fixes.

Fingerprint loops, tool latency, context pressure, large params, rare tools, and side-by-side diffs.

Use it locally, then gate it in CI.

The same parser powers the TUI, Markdown and HTML reports, JSON output, GitHub artifacts, and threshold-based build failures. 

agenttrace --overview -f html -o agenttrace-overview.html
agenttrace --overview -f markdown -o agenttrace-overview.md
agenttrace --overview \
  --fail-under-health 80 \
  --fail-on-critical \
  --max-tool-fail-rate 15

Multi-agent logs, one operational view.

Claude Code · Codex CLI · Gemini CLI · Qwen Code · Aider · Cursor exports · Hermes Agent · OpenCode · OpenClaw · Kimi CLI · Copilot-style traces

Built in the open for agent operators.

agenttrace is being submitted to terminal, AI agent, observability, and LLMOps directories. Feedback is tracked publicly so the TUI can keep following real coding-agent pain.

Star the repo Share feedback Launch kit Release notes