wtf is minelogue?

Minelogue is a tool for looking through the log files left behind by AI coding agent sessions, specifically Claude Code, which stores logs as JSONL files, and OpenCode, which stores them in SQLite. Instead of scrolling through raw log files, Minelogue turns a session into an interactive timeline and transcript you can browse in a web dashboard, including the many separate subagent conversations that can branch off from a main session and how they were spawned. The name is a play on words, combining mine as in extracting something valuable with the ending of words like dialogue and monologue, plus a nod to log files. The stated roadmap is to first make every session's conversation easy to browse, which is what it does today, then add the ability to cross examine what shows up in a session, and eventually support having an AI agent judge and grade other agent sessions. You run it as a local Python application, pointing it at your Claude Code and OpenCode data folders, and it opens a dashboard in your browser. A lot of engineering effort has gone into making this fast even for very large sessions: instead of loading full message contents up front, it loads lightweight previews first and only pulls in full details for a message when you actually need to see it, and it keeps a local cache so that reopening a session you have already viewed is close to instant. The interface includes clearly labeled elements meant for automated browser testing, and the project ships its own test scripts that build sample sessions and check that the dashboard renders correctly. The project also includes standalone utility scripts for scanning Claude Code log files to understand the different kinds of messages and structures that appear in them, useful if you are building your own tooling around these logs. Minelogue uses a dual license setup. The core source code is released under the AGPL 3.0 license, and a separate commercial license is available for anyone who wants to use it in closed source or proprietary products where the AGPL terms would not fit.