Codex Trace CLI
Codex Trace reads local Codex CLI logs and shows turns, tool calls and token usage in a desktop window or in the browser.
- 120 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Installation runs a script fetched with curl, and session logs may contain code and secrets.
Install
- Works in
Manual install
curl -fsSL https://raw.githubusercontent.com/PixelPaw-Labs/codex-trace/main/script/install-macos.sh | bashInstalls `Codex Trace.app` into `/Applications`, Apple Silicon only. For Linux and Windows use the builds from the releases.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Installation runs a script fetched with curl, and session logs may contain code and secrets.
This is third-party code. Review the repository files before installing.
What it does
The app opens session files from `~/.codex/sessions/` and turns raw JSONL into a readable conversation. A date-grouped session tree sits on the left, turns in the middle and turn details on the right. It shows commands, MCP tool calls, patches, web searches, image generation events and links between orchestrator and worker sessions. It offers search across sessions and messages, live tailing of a running session and token counts where recorded. It runs as a Tauri desktop app or a web interface, including in Docker. Nothing is sent anywhere; everything is read from disk.
Who it is for. Developers running long Codex CLI sessions who need to review what the agent did, where tokens went and why a step broke.
Good fit when
- A Codex session went off track and you want the turn where it began.
- You want to see which commands and patches the agent ran in an evening.
- An orchestrator spawns worker sessions and their links need a full view.
- A running session is watched from a browser on a second screen.
Not a fit when
- You need logs from Claude Code: claude-code-trace covers it.
- You want a cloud observability platform with team features.
- The Codex CLI log format changed after the supported versions.
Example request
Open the latest Codex session in Codex Trace and find the turn where the agent started editing the wrong files.Limitations
It reads only Codex CLI JSONL logs; versions from 0.44, a mid format and the August 2025 format are supported. The macOS one-liner downloads the app via a script and targets Apple Silicon; take Linux and Windows builds from the releases. Token counts appear only where Codex recorded them. Logs may hold code, paths and secrets from sessions, so keep web mode bound to `127.0.0.1`. Last commit on 31 October 2026.
How to disable. Remove `Codex Trace.app` from `/Applications` or stop the container. The program does not modify Codex logs in `~/.codex/sessions/`.
FAQ
Where does the app get its data?
From `~/.codex/sessions/YYYY/MM/DD/rollout-...jsonl`. The folder can be changed in settings with the `,` key.
Can it run without installing the app?
Yes. `npm run dev:web` from a clone opens the interface in a browser, and Docker runs web mode only on port 1422.
Does it see Claude Code logs?
No. The README points to a separate project, claude-code-trace, for those.
Is data sent out?
No. Per the README everything is read locally and traces do not go to an external service.
Which systems does it support?
macOS, Linux and Windows. The one-line install targets macOS on Apple Silicon; others need builds from the releases.
Discussion
No questions or reviews yet.
Related
CLI toolsA Codex plugin that turns local task logs into a timeline with token counts, tool timing and estimated cost for every step.
Reads Claude Code session logs as a conversation: tool calls, token counts and a live tail of a running session.
Codex MCP Tool (cexll)
Codex MCP Tool
An MCP server that sends tasks to Codex CLI from Claude Code and Cursor: code review via @ refs, a sandbox, patch edits, ideas
Local web panel for Claude Code and Codex sessions: one sidebar, real terminals in browser tabs, and cost tracking.
A Go CLI that gathers Claude Code, Codex and Copilot sessions into one searchable base with a TUI and an MCP memory server
Reads local logs of Claude Code, Codex, Cursor and other agents and shows token spend, time and failures in a TUI
README in short
The tool targets local Codex CLI logs and needs no external trace service, unlike general observability platforms.