All Pro features are free for a limited time — workflows, the AI backlog (PO loop), scheduled jobs, screen mirroring, terminal, local inference, git worktrees and live preview are fully unlocked, no purchase needed. Security hardening — file protection & backup exclusion for onion auth keys, a tighter network-access policy, and masked IPs/addresses in diagnostic logs for stronger privacy and supply-chain safety.

Repeat runs — re-run a task automatically «until it passes». Each attempt starts in a fresh session (clean context), and it stops when the check passes or hits your max-tries limit. Start one without building a workflow canvas. Check command — instead of the AI judging itself «done», completion is decided by the exit code of a check command you choose (tests, lint, type-check, build). It only finishes when that actually passes. Loop monitor — see the iteration count, the last check's pass/fail and reason, and a change summary on one screen — and stop the loop with a single tap. Scheduled-collection alerts — your daily backlog collection now tells you, by notification and card, whether it found new proposals, came up empty, or failed midway. Runs ending with no real result are flagged «attention», not «done». Reliability — fixed a Face ID lock that could get stuck re-prompting, and blocked a «zombie» connection that looked connected but was unresponsive.

Live Activity — your agent fleet's status now shows on the Lock Screen and Dynamic Island, so you can see who's working, waiting, or done without opening the app. Connection diagnostics — a new read-only screen shows the health of each subsystem (SSH, Tor, daemon), so you can tell at a glance why a connection is off. Problem reporting — if the app hits a problem, it keeps a local log you can review and share from the new diagnostics screen. Backlog — the AI product-owner loop now runs a self-correcting workflow (draft → check → retry), gathers evidence in multiple passes, and renders briefs as readable Markdown with a one-glance triage summary. Reliability — hardened phone↔Mac reconnection so sessions recover cleanly after a drop.

Backlog AI — now gathers evidence from 11 source types through expert lenses (security, design, and more), and each suggestion shows which expert reviewed it. Localization — the backlog's collection, research, and revision results now appear in your device's language. MCP tools — a live probe verifies each connection for a more accurate “connected” status, removing false greens. Sessions — when terminal input fails to send or drops, you now see it right on the session screen. Stability — “Stop” now interrupts work in Antigravity, OpenCode, and local-LLM sessions, and we fixed lost workflow runs, briefs, and terminal output.

Same-Wi-Fi-only connection mode — connect directly only on the same Wi-Fi with no outside network, blocking everything else. On first launch you choose how to connect (anywhere vs. same-Wi-Fi-only). Backlog (PO loop) triage — sort the approval queue by score tiers and filters; the report card breaks down rejection & verification reasons plus post-release misses. Workflow run history — view recent runs on your phone, tap to open or re-run them, and get status transitions as notifications. Added GitHub Copilot CLI as a built-in agent. Screen mirroring gains one-finger scroll and circle/arrow markup; archive sessions, group by repo, and clean up in bulk.

Install a local LLM (llama.cpp · Qwen Code) on your Mac in one tap from your phone, and run sessions on local models — no cloud. Run sessions through the new OpenCode adapter — or point it at your own OpenAI-compatible server, like Ollama. Describe it in one sentence and AI drafts a multi-agent workflow for you. A checkpoint timeline rolls you back to the last safe point in one tap. SSH host-key verification, Tor bridges, and a Security status panel keep connections safer — and if something's blocked, a card tells you what and why. Install missing agent CLIs onto your Mac straight from your phone, and the terminal width (auto / custom / wide) is remembered per agent.

Almost every input now has a mic — dictate session names, searches, the remote keyboard, and commands by voice. In live preview, mark up the screen with your finger and send feedback that even identifies the element you pointed at. Settings now has Device management — see paired devices, allow a second device, or revoke them individually. Share a session preview via the system share sheet, and pick a mirroring capture target faster with per-app groups and search.

Live-preview your dev server right on your phone — capture the screen, mark it up, and send it to chat as feedback. Your session list is now grouped into Running, Waiting, and Done, so you can orchestrate many agents from one board. Scheduled and recurring tasks now live in a new Automation tab. Find across chat, files, and your session list, and jump between diff hunks. Plus expanded voice dictation and reply previews in notifications.

Connect two iPhones to one Mac — allow a second device under Settings ▸ Devices and use both at once. See all your projects at a glance — filter the backlog by repo and track progress with the cumulative scorecard. App Store reviews and post-release crashes now feed product decisions, so your next tasks get smarter. Sessions are cleaned up automatically when research finishes, and in-progress items can be swiped away.

Choose which agent runs your backlog work — pick a CLI tool before starting signal collection, research, or an approved brief's implementation. In-progress research items can now be swiped away. Stability improvements and minor bug fixes.

New: Backlog — the app gathers improvement signals, drafts a brief, and once you approve on your phone, implementation begins. Annotate attached images with pen and blur tools — redact sensitive info before sharing. The diff viewer now features syntax highlighting for easier code reading. Mirroring: capture just the window you choose, remember your last zoom position, and optionally stream only the zoomed region. You can now adjust your local LLM's context size in Settings.

Introducing Pocket Sisyphus Pro. The core app stays free. Features marked “Pro” — workflows, scheduled tasks, screen mirroring, advanced chat tools, and Terminal & local-LLM sessions — unlock with a monthly or yearly subscription, or a one-time lifetime purchase. Subscriptions include a 7-day free trial. We also keep your Mac’s screen from sleeping during mirroring and polished overall stability.

• Live screen mirroring — watch your Mac on your phone, pinch to zoom, and control it remotely (tap, scroll, drag, custom shortcuts). • Artifact viewer — browse and preview files your agents produce. • Smoother streaming and many fixes.

• Workflows — chain multiple coding agents as nodes and run them in one tap • Scheduled tasks — run agents automatically on a cron schedule • Added an in-app guide for workflows and Pro features • Connection stability and various refinements