Hi Product Hunt! I've been writing software for 17+ years, so when I started testing Claude Code locally, then opencode and a few others, I wanted to know what they were all actually doing under the hood. Which files they read, what they send out, what state they leave behind. Each one keeps its session state in its own format, in its own directory, with its own idea of what a "session" is, and there was no way to look across them. So I started writing something to read those files, and it turned into threadle. Your agents already write everything down: transcripts, tool calls, token counts, the files they touched. Almost nobody reads it. threadle reads it and draws it. Every session becomes a node graph you can walk, diff and trace across runs. And once they're all in one place, you can hand context from one tool to another and wire sessions into pipelines. It all runs on 127.0.0.1. No account, no telemetry, no outbound requests. Free and MIT licensed. npx threadle@latest I'd love to hear which agents you use, what you'd want to see in your own sessions, and what breaks. Comments here or GitHub issues both work. Thanks!

threadle

What your coding agents did while you weren't looking

What your coding agents did while you weren't looking