Your phone now keeps its encryption key when Pushary is reinstalled or secure storage is briefly unavailable, so saved conversations stay readable. If history is locked on a new phone, the workspace owner can choose Use this phone for history in Settings, then verify this phone on each computer that saves history. New conversations then open on this phone; conversations saved for the other phone stay locked. Settings also shows this phone's key fingerprint. Conversations saved with more than one key now show every message this phone can read, and a question no longer shows in other conversations from the same agent.

Claude and ChatGPT questions now show their own logos in your inbox. Risk warnings remain separate so you can identify both the agent and anything that needs attention. Set up Claude or ChatGPT from Settings. Claude setup opens the published Pushary directory listing, with manual connector setup available if needed. Connections recover more smoothly when several sessions reconnect at once. Follow live replies from compatible Claude Code sessions opened on your Mac. Conversations recover more reliably after a disconnect, and replies stay in place as they finish, including while Claude is thinking. A message you send while Claude is working now shows in the conversation as yours once Claude picks it up. Instructions and skills an agent loads, such as CLAUDE.md, show as a short References row you can open instead of filling the conversation. Agents that connect through MCP, such as fx, now show their questions on their session screen so you can answer them there. Every agent shows its own logo or initial instead of a question mark. When session history is off for your workspace, or this phone is still being set up, the session screen now says so. With an updated helper on your computer, supported Gemini, Cursor and Hermes sessions can now sync their displayed conversation to your phone. Edits and rewinds update the saved conversation, and capture notices explain when history is unavailable or limited. Stale reply previews disappear when the conversation moves on. Workspace history choices apply consistently across agents and sites. After you have approved three requests, Pushary may ask once for an App Store rating. It only asks when your inbox is clear, never while a request is waiting for you.

Claude and ChatGPT questions now show their own logos in your inbox. Risk warnings remain separate so you can identify both the agent and anything that needs attention. Set up Claude or ChatGPT from Settings. Claude setup opens the published Pushary directory listing, with manual connector setup available if needed. Connections recover more smoothly when several sessions reconnect at once. Follow live replies from compatible Claude Code sessions opened on your Mac. Conversations recover more reliably after a disconnect, and replies stay in place as they finish, including while Claude is thinking. A message you send while Claude is working now shows in the conversation as yours once Claude picks it up. Instructions and skills an agent loads, such as CLAUDE.md, show as a short References row you can open instead of filling the conversation. Agents that connect through MCP, such as fx, now show their questions on their session screen so you can answer them there. Every agent shows its own logo or initial instead of a question mark. When session history is off for your workspace, or this phone is still being set up, the session screen now says so. With an updated helper on your computer, supported Gemini, Cursor and Hermes sessions can now sync their displayed conversation to your phone. Edits and rewinds update the saved conversation, and capture notices explain when history is unavailable or limited. Stale reply previews disappear when the conversation moves on. Workspace history choices apply consistently across agents and sites.

Clearer agent conversations with formatted replies and smoother streaming. Improved phone session starts, message recovery after reconnecting, and session navigation.

Turned off renewal in the App Store or Google Play but still using Pushary? Your Inbox now shows when your plan ends, with one button to turn renewal back on.

When a session shows only a summary, the app tells you why. Invited by a company? Scan your invite is now a full button on the welcome screen. After you connect, one button opens your requests. If the company can't connect new people right now, the app says so plainly. Companies that use Pushary with their own customers can send themselves a test approval to see what their customers see.

After you pair from your terminal, Done opens your inbox, where the setup test question arrives. Tapping the notification for a test question opens that question. Connecting from Pushary for Mac takes two fewer screens to reach your inbox. Answers you did not send are listed under Unsent answers, below your waiting questions, and no longer show while you are still answering.

After you connect a computer, Done takes you straight to your inbox. Pushary asks for notifications once. If you turn them on in Settings, the app picks that up when you come back. Resume approvals from a paused session, even after opening or dismissing the notification. Answers you start typing are saved on your phone for up to 24 hours under Saved drafts. Choose how questions reach you with shorter, clearer modes: When away, Always, Alert only, and In agent. Company invite recipients can connect their Mac and receive company approvals there. This build also prepares phone sign-in with matching codes for a compatible Mac release. Existing sign-in and QR pairing remain available while that rollout is disabled.

Try it now checks the computer you just set up, and names it on the screen. Signing in on a new device asks for the code we email you, instead of saying sign-in failed. Start a task explains how to start sessions from your phone on a Mac: open Settings, Phone, and keep the Mac awake with Pushary running. It also tells you when a computer is connected but not ready yet. New session stays in Fleet even when no computer can start a session yet, so the setup steps are one tap away.

While a connected computer waits for its first agent, the app shows a row with the Claude Code, Codex and Cursor marks where the agent will appear, instead of a dot. Once a computer is connected, Share setup link stays the one main button. If the computer found no agents, the setup command is shown so you can run it again. The opening line on the subscription screen is easier to read.

Connecting your computer now shows one clear next step on every screen: share the setup link, get a new link, or try your agent. The terminal steps show the command in a box, the command for Windows and the Node.js versions setup needs. If no computer connects after two minutes, the app reminds you to use the same account on your computer and this phone. The finish screen is simpler, and shows the one step any agent still needs. Signing in on a new phone asks only to turn on notifications when your agents are already connected.

Clearer steps to connect your computer and try your first agent question. Scan a setup QR directly from onboarding, or open terminal instructions for Mac, Windows and Linux. Improved setup recovery and clearer sharing instructions.

Scanning the setup QR from your terminal shows which computer is asking to connect, its operating system, and the agents setup found. Deny stops setup on that computer right away. A QR scanned before signing in or subscribing opens after you finish setup. Clearer onboarding instructions and improved sign-in buttons. Analytics settings on Android now describe only the data shared on Android.

The setup screen can now send you a test approval and wait for your answer there. Scanning a pairing QR or opening a notification while you are still setting up now takes you to it, instead of back to setup. Pairing a computer to a workspace with no active subscription now says so, and asks you to subscribe in Settings first. The plans now say what each one gives you: Agent covers every agent you run, and Agent Pro adds unlimited notifications, budgets and priority support. Deleting your account now reminds you to cancel any App Store subscription first, with a button that opens your subscriptions. Status labels have a softer look that is easier to read. Opening the app without a connection no longer shows the subscription screen to someone who already subscribes. If you subscribe on pushary.com while that screen is open, the app moves on by itself when you come back to it. Stop and Send on a session are their full size again. Signing out now clears the previous account's approvals and sessions from the screen. If you turned analytics on in an earlier version, the app asks once more, because it now also reports whether an Apple Ads campaign brought you to the App Store.

Deleting your account now asks why first. Answering is optional, and if the reason is that Pushary asked you to approve too much, it points you at the Policies tab, where rules can cut that down. After you approve Bash three times in a row in one agent session, the approval card offers to allow Bash for the rest of that session. Risky commands, like deletes, pushes and network calls, still ask. A rule you save from an approval card is written only once your answer is recorded, and if it cannot be saved the card says so. A session rule for a command now names what it covers, such as all "npm run" commands. Audit now marks decisions a rule or a timeout settled, so they no longer look like ones you tapped. And where the law requires consent for usage analytics, you may be asked once more, because earlier versions did not record whether an answer came from you.

The usage analytics question is shorter and clearer. Two lines instead of five, with the full detail behind a single tap rather than a wall of text, and the two answers side by side at equal weight. Where local law asks for notice rather than a prompt, it no longer interrupts first launch. The switch still lives in Settings, Privacy, and where consent is required nothing is measured until you allow it. Also repairs a hold left behind by an older command line version, and reads the blocked action marker correctly on the phone.

Nothing inside the app changed in this version. It carries a new set of App Store screenshots, which only reach the store attached to a version. The screenshots now show the screens you actually use: an approval as it arrives, the intent and the exact command behind it, the rules that decide what runs on its own, your agents on one screen, and the record of every decision.

A set of questions from an agent now comes one at a time. A strip at the top shows how many there are and which you have answered, picking an option moves you on, and a review step lists every answer before you submit. The subscription and upgrade screens were rebuilt. Both plans and their prices are visible without scrolling, the free trial length comes from the store rather than being written into the app, and Restore purchases is offered even when the plans cannot be loaded. A question waiting on you now sits at the end of the conversation it came from, and the screen scrolls to it. On the other tabs it sits at the top, where a long list used to hide it. The inbox no longer offers a decision that is being handled in your terminal, and each question shows the mode that actually applies to it. Agent replies now use the app's own typeface throughout, including bold text, smaller headings and tables, which previously fell back to the system font. Text you type sits on a white field, and the audit list shows that it is loading instead of looking empty.

Each account on this phone now has its own answer to the usage analytics question. When another account signs in, it is asked separately, and nothing is shared for it until it answers. Fixed being stuck on "Couldn't load your account" after switching accounts while the store could not be reached. You now reach the plans screen, where you can retry or sign out. If a subscription is charged but confirming it fails, the purchase stays linked to your account. While your account is still loading, the upgrade screen now says so and offers Retry account, the same as the plans screen.

A request you were told about now shows up in your list even when the list has not caught up with it yet, and it stays there until you answer it. Before this, a request could arrive on your lock screen and be missing from the app when you opened it. The list no longer blanks out and reloads while you are reading it. It refreshes quietly in the background, and only says it is loading the first time you open the screen.

Conversations from sessions you start in your own terminal now reach this phone. Before this, only sessions launched through Pushary synced their history, so a session you started yourself showed a summary instead of what was said. Replies render as formatted text. Headings, lists, bold, links, tables and code blocks, instead of the raw Markdown punctuation that used to show through. A reply now appears while it is still being written, marked Live reply until it finishes. A code block or table that has not closed yet is held back rather than drawn half written. Sending a message no longer takes the terminal away from you. Messages wait for the agent's next turn, and Continue on phone is a separate action you choose. If the terminal cannot hand over, it now tells you why instead of going quiet. The audit list shows a short outcome and the answer underneath, so a reply to several questions reads as questions and answers rather than raw JSON. When history cannot be loaded or decrypted, the screen says so and offers a retry, instead of falling back to a summary without explaining.

Synced conversations are now described accurately. The app used to call them end-to-end encrypted. A conversation is encrypted on your computer before it is uploaded and this phone can read it, but authorized Pushary compliance admins can also recover one exact session under a documented legal obligation. The session screen and the transcript section now say so. The command that turns syncing off sits on its own line and is selectable, so you can copy it instead of retyping it.

Handing a decision back to your machine now works. Tapping "Handle on your machine" used to fail silently, leaving the card on screen and the agent waiting; it now ends the decision and hands control to your terminal. An agent that has gone quiet is no longer reported as unreachable. The message box and Stop stay available on a session that is simply waiting at its prompt, and the reason is accurate when they are not. "Working" now clears when the work is done, instead of staying lit under a finished task.

An approval now tells you how long you have. Questions expire, and the card counts down so you can see at a glance which one needs you first. The approval itself is simpler. Approve and Deny are the only choices in front of you, and the standing rules that stop Pushary asking again are one tap away instead of stacked underneath. A session reads as a conversation. What you asked and what the agent answered sit in one thread, with code shown as code and a message box pinned at the bottom so you can reply without hunting for it. Sessions that failed no longer vanish when you hide finished work, and your decision history says what happened in plain words rather than status codes.

Your agent sessions now read as a conversation. What you asked and what the agent answered sit in one thread, in full, with a message box pinned at the bottom so you can reply without hunting for it. Starting a session no longer asks you to type a folder path on a phone keyboard. Pushary remembers the folders you have used on each machine and offers them as a list you tap. If you are on the Agent plan, the screens that need Agent Pro now show you what you are missing and let you upgrade from your phone, instead of stopping at a message.