You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
You can use specstory resume to resume Muse Code sessions in other coding agents, and to resume any agent's session into Muse Code.
🐛 Bug Fixes
output_dir in config.toml is now honored by specstory watch and specstory resume, not just by run and sync. Previously only the --output-dir flag reached watch, so with the setting in your config file the startup save went to your configured directory while everything written during the watch went to ./.specstory/history — leaving markdown in two places, with the configured copy going stale. Thanks to petercurcio for the detailed report in issue 273. If you hit this, the copies under ./.specstory/history are the newer ones; move or delete them once, and everything lands in your configured directory from now on.
Codex CLI sessions started in Codex 0.147.0+ are saved again. Codex changed how its Codex CLI TUI records a conversation in that release, and SpecStory read only the older form, so those sessions produced a markdown file containing just a header, with every prompt and reply missing. Both forms are now read. This affects only sessions started in 0.147.0+. Sessions from earlier versions, and older sessions you resume under 0.147 were never affected, and their markdown is unchanged. Re-run specstory sync to fill in any sessions that were saved empty.
specstory watch and specstory run now behave the same way at startup whichever agent you are watching: they save new activity from that point on, and leave sessions that already exist alone. Previously Codex CLI, Droid CLI, DeepSeek TUI and Antigravity CLI re-saved and re-synced their existing sessions on every start, while Claude Code, Cursor, VS Code Copilot, Gemini CLI and Muse Code did not. Run specstory sync when you want existing sessions written or refreshed.
The first update to a session that already had a markdown file is no longer missing from specstory watch output. The line was sometimes being swallowed as if it were startup noise, so for some agents, some of the time your first prompt after starting watch appeared to do nothing; only the second one showed up.
specstory run and specstory watch no longer reprocess your entire Codex CLI history every time they start. Codex stores sessions by date, and while SpecStory only ever watched the last few days for live changes, it was scanning and re-saving every day directory going back to your first ever session — rewriting all that markdown and re-syncing it to the SpecStory Cloud on each launch. Startup now covers the same trailing window that live watching does, plus the day of a session you are resuming, however old it is. If you have a long Codex history, expect run and watch to start noticeably faster and stop touching old markdown files; a full specstory sync still processes everything as before.