Hey Product Hunt 👋 I kept ending up with multiple Codex sessions for the same project, often starting with the same handoff prompts. Once I had dozens of them, figuring out which session to resume got kinda confusing. I sometimes ended up guessing from the agent's last reply or token count because the context I actually needed was buried inside the transcript in raw JSONL. Another problem was reading old sessions. I wanted much finer control over what I was looking at: assistant final/progress/reasoning vs tool calls/output/hooks and everything else. I was literally copying useful explanations into Notepad as Codex worked — sometimes keeping the terminal open just so I could catch those mid-work explanations and paste them somewhere to read properly later. I wanted to keep up with what the agent was doing so I could guide it better and not lose the plot as projects got more complex. Eventually I constantly had 10–15 Notepad tabs open. That just created another mess. So I built **Agent Session Browser**. It's a local web interface that lets me browse my Codex CLI history by project/date, actually read the transcript before deciding which session to resume, choose which reasoning/tool/progress/provider events I want to see, export sessions, and resume the right one. There's also a TUI for when I'm already in the terminal: session list on one side, actual transcript on the other. It also supports Claude Code, Gemini CLI, Pi, and Antigravity. Fully local, read only, and MIT licensed. Run it: npx agent-session-browser Would love to hear your feedback

Agent Session Browser

GUI + TUI for browsing coding agent session history

GUI + TUI for browsing coding agent session history