Free up disk space by finding and safely deleting old Codex AI conversations you no longer need.
Search through past Codex sessions by keyword to locate a specific conversation.
Clean up leftover orphaned database entries after Codex conversations are removed, and see how much space will be recovered before confirming.
codex-history is a command-line tool for browsing and deleting conversation history that the Codex AI coding tool stores locally on your computer. Codex saves past conversations as files on your machine, and over time these can pile up and take up disk space. This tool gives you a safe way to find specific conversations and remove them without accidentally deleting the wrong things. The main commands let you list your saved conversations, search them by keyword, and delete one or more by ID. Before any deletion happens, the tool shows you exactly what it found and asks you to confirm by typing a short code. If any of the targets you asked to delete cannot be uniquely identified or appear to be actively in use, the whole operation is cancelled rather than proceeding with a partial delete. There is also a cleanup command for orphaned data, which refers to leftover database entries and log records that no longer have a matching conversation file. The tool calculates how much disk space this orphan data is taking up and shows you the plan before asking for confirmation. A doctor command checks whether the current version of the tool understands the data format that your installed version of Codex is using. Since Codex updates can change the local file structure, running doctor after a Codex upgrade is recommended before doing any deletions. The tool works on macOS, Windows, and Linux. It only touches local files on your machine and has no effect on any server-side Codex history. After deleting conversations, restarting Codex Desktop is recommended so it picks up the changes from disk. The project is MIT licensed.
← liuyoumi on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.