v0.24.2

[0.24.2] - 2026-10-21

Amp orb sessions now reach your team whole and on their own, Oxy can join you in Claude Code, and plan reviews, approvals and saves reliably reach your team.

New

• Amp orb sessions reach your team on their own — in an Amp orb, ox records each thread and uploads it on its own, credited to the owner of SAGEOX_TOKEN. Every Amp session now starts with the prompt that opened its thread, and a reply sent after the orb went idle is recorded as a continued session.
• Record Cursor Agents Window sessions (preview) — ox integrate install --cursor captures them and shares them with your team, like Claude Code sessions.
• Hatch Oxy in Claude Code — ox pet enable puts Oxy, SageOx's mascot, in Claude Code's thinking line, where he reacts to what Claude is doing: swatting a bug during a test run, launching a rocket when you push, throwing confetti when a turn ends. /hatch-pet shows, hides or resizes him.
• Draw around part of a plan and comment on exactly that — in review mode, loop around a paragraph, table rows or diagram nodes, and your comment stays attached to what you circled even after the plan changes.
• Comment on a plan as a whole — in review mode, the comments panel takes a comment on a plan's direction, scope or readiness without pinning it to a section. It reaches the AI coworker like any other comment.
• Your AI coworker reads screen walkthroughs moment by moment — ox conversation walkthrough pairs each click, selection and pause with what was said and what the screen did next, shows a close-up of the element on request, and splits long recordings into chapters.
• ox config knows every setting — ox config schema --json lists each setting with its description, accepted values, levels and current value, even outside a repository, and ox config --help now lists exactly the settings ox config set accepts.

Improved

• New team discussions reach your AI coworker sooner — they arrive on its next check instead of up to five minutes later.
• ox index github keeps the whole conversation — imported issues and pull requests keep every comment and the right draft or closed state, and they're searchable as soon as the command finishes.

Fixed

• Plan reviews, approvals and saves reliably reach your team — each is sent once and retried until your team's Ledger holds it, even through dropped connections, reloads and two people acting on one plan at once.
• Approving a plan or ending a session never publishes your unsaved edits — only the plan's status and history are shared.
• Sync keeps going through conflicts and interruptions — a plan changed on two machines, an interrupted push, or a save during background sync no longer leaves your Ledger stuck. While a repair is needed, ox writes nothing and points you to ox doctor --fix.
• Team memory notes save while Team Context syncs — ox memory put and the note ox agent session stop records wait for the sync instead of failing.
• Plan review is more dependable — Review responds to your first click, a plan's own tabs and checkboxes work, marks on Mermaid diagrams stay where you put them, and the review page and ox plan commands flag damaged or conflicting feedback instead of showing a partial review.
• ox logout reliably signs you out — a background token refresh can no longer bring the login back, expired logins are cleared too, and a newer login it didn't revoke is never deleted.
• Large Ledgers finish their first download — it's no longer cut off after a minute, so teams with a long history get their Ledger instead of retrying forever.
• A session you abort stays private — ox session import no longer offers it, and ox import says a file is imported only once the server has accepted it.

Security

• Credentials stay out of shared sessions — every recorded or imported session is checked before it reaches your team, and anything that can't be checked is held back.