Orca v1.4.242 opens Claude subagents from the sidebar, and Codex Ask mode asks before it edits

58 minutes ago

Orca v1.4.242: click a Claude subagent in the sidebar to open its transcript, and Codex chats in Ask mode now ask before they edit your workspace. Also: resumed agents keep their effort level, flags and permission, paired servers keep your tabs across restarts, and terminals use far less idle CPU. Release notes: https://github.com/stablyai/orca/releases/tag/v1.4.242 Compare: https://github.com/stablyai/orca/compare/v1.4.241...v1.4.242 Download: https://onorca.dev/download

Ask about this presentation

Chapters

  1. 0:00Claude subagents in the sidebar
  2. 0:12Codex Ask mode asks first
  3. 0:25Agents resume as you left them
  4. 0:37Remote servers that remember
  5. 0:49Lighter on memory
  6. 1:04Three fixes
Show transcript
  1. Click a Claude subagent in Orca's sidebar or dashboard, and its transcript opens in the same terminal. Click the parent to go back. Claude sessions already running need a fresh launch.

    Slide text
    Orca

    Click a Claude subagent in Orca’s sidebar to open its transcript

    It opens in the same terminal. Click the parent to go back, and an eye marks the one you’re viewing.

    Default
    on, with Claude status hooks
    Shortcut
    Orca adds one key to Claude’s keybindings.json
    Running sessions
    need a fresh launch
    stablyai/orca release notes on GitHub · PR #33322
  2. A Codex chat in Ask mode could edit your files without asking. Now it runs read-only and asks before each edit. Chats on a paired server wait until that server updates.

    Slide text
    Orca

    Codex chats in Ask mode now ask before they edit your workspace

    Ask runs read-only until you allow the edit. Chats on a paired server keep the old behavior until that server updates.

    A Codex chat in Orca about to run apply_patch, with an Apply file changes? prompt offering Allow, Allow for this session, Deny and Stop
    stablyai/orca release notes on GitHub · PR #31299
  3. Resume a Claude session and it keeps its effort level and launch flags. A resumed Codex thread keeps the permission you last gave it, instead of falling back to Orca's launch flag.

    Slide text
    Orca

    Resumed agents keep the settings you gave them

    Claude
    resumes at its own effort level
    Background launches
    keep their launch flags
    Codex
    resumes with the permission it was last given
    stablyai/orca release notes on GitHub · PR #33765, #33776, #33460
  4. Restart Orca, and a paired server's tabs come back, while tabs you closed stay closed. WSL hosts stop asking to pair again, and a failed SSH connect shows what SSH said.

    Slide text
    Orca

    Paired servers keep your tabs across restarts

    Closed tabs
    stay closed after a restart
    WSL
    hosts stop re-pairing after each restart
    SSH
    a failed Connect shows what SSH said
    stablyai/orca release notes on GitHub · PR #33926, #33702, #33109, #33928
  5. Orca now checks idle shells less often, so terminals burn far less CPU while you wait. The terminal host shrinks once it next restarts, and WSL connects through bash instead of Node.

    Slide text
    Orca

    Terminals use far less idle CPU

    Terminal host
    uses less memory once it next restarts
    WSL
    connects through bash instead of Node, with far less memory
    Resource panel
    matches Activity Monitor on macOS
    stablyai/orca release notes on GitHub · PR #33829, #33837, #33820, #33901
  6. Windows launchers now run from folders with spaces, Korean input stops sending word jumps twice, and automations survive daylight saving. If you run Claude subagents, I'd update today: one click opens each one.

    Slide text
    Orca

    Three fixes you may have hit

    Windows
    launchers run from folders with spaces
    Korean input
    word jumps send once while composing
    Automations
    no skipped dates around daylight saving

    Run Claude subagents? Update today.

    Orca v1.4.242 release notes on GitHub

    stablyai/orca release notes on GitHub · PR #13501, #17624, #19086

What happened in AI today, in your inbox

One short email each evening. Read it and you’re caught up.

Built in New York with ❤️