Ask User The assistant can now ask you questions or request a confirmation on a card in the chat, wait for your answer, and resume afterwards. Answer multi-question cards one step at a time, by voice, or straight from a notification when Minis is in the background. It never decides for you: there's no "Always allow," and a timeout is reported back as unconfirmed. Share chat as a long image or PDF Share any part of a conversation as one long image or a PDF. Pick individual replies, tool calls, and thinking with drag-select, fold or hide tool runs, choose phone, iPad, or adaptive table width, follow light or dark appearance, and zoom the preview. The header shows the assistant's avatar, the model, and the date. Search in Chat Find text in the open conversation from the "…" menu or a long-press, step through the results, and see matches highlighted in place. Session search also supports several keywords and shows the matching line. Scheduled tasks you can see A small pill under the navigation bar counts down to the next scheduled run, shows which run is next, and lets you run it now, skip it, or cancel. A conversation waiting for its next run also keeps a card in the Live Activity. Each scheduled task has its own detail page with parameters, status, and run history. Quota and balance at a glance See your remaining quota or balance for supported providers, including Claude and ChatGPT subscription limits, in the model picker, on the provider page, and in the input bar. The model dot turns yellow when it runs low. Fold consecutive tool calls Turn it on in Settings → Appearance → Tool Calls to fold a run of tool calls into one "Ran N tools" line with the total time. Faster, steadier background runs On iOS 26+, a continued-processing task keeps long runs going in the background, with a card that shows what the run is doing. Always On mode keeps the screen awake with a dimmed black screen saver. AOJIT acceleration Common built-in tools now run from precompiled execution images, so they start and finish faster. It's on by default; if anything misbehaves, turn it off in Settings → Rootfs Management → AOJIT. More ways to control your data Disable environment variables without deleting them. Clean up orphaned session files in Settings → Storage. Enable or disable all skills and MCPs in a session at once. Long-press a skill to copy its source URL. Voice Send waits for speech recognition and offers a retry or a switch to the next voice model if it stalls. Voice correction restores English terms and uses what's on screen. A volume slider is available for read-aloud. New models Added Claude Sonnet 5.5, GPT-6.1 Sol, and GPT-6 Astra. Bug Fixes - Fixed long conversations running into "Context is full" more often than necessary; old reasoning and large outputs are now trimmed before any compaction, and context size is estimated more accurately - Fixed a launch hang of a few seconds on large histories, and the home list rebuilding every row - Fixed several crashes: sharing a long image and backgrounding the app, a denied OAuth sign-in, undo in the input box, iPad keyboard switching, and editing a context window on iPad - Fixed the app freezing on a mounted iCloud folder, and Stop not ending a tool whose process can't exit - Fixed background keep-alive stopping after a phone call - Fixed backups: a server backup that couldn't be restored, SMB logins failing at random, and a deleted model group not coming back - Fixed scrolling jitter in long chats, rows overlapping while a reply finishes, and the list jumping after rotation - Fixed tool-call arguments being lost for Anthropic and Responses API models, and images not reaching Anthropic on API-key routes - Fixed a draft in a new chat disappearing after going back

Subagents Minis can now delegate parts of a task to sub-agents that work in the background while you keep chatting. You can watch their progress inline, open a full transcript of what a sub-agent did, steer it mid-task, or stop it — and multiple sub-agents can run and queue up in the same conversation. This capability is now enabled by default. Scheduled tasks in chat You can now schedule a prompt to run later directly from a conversation. Scheduled runs show up as a card in the chat with their next firing time, and you can cancel a pending one right from that card. Context usage indicator in the composer The input box now shows how much of the conversation's context window is left, with a subtle edge glow that warns you as a conversation gets close to its limit — so you can see it coming before a reply gets compacted or cut off. More languages supported Added Indonesian, Malay, Filipino, Thai, Turkish, Polish, Romanian, and Croatian, plus translation gap fills in German, French, Russian, Japanese, Korean, Traditional Chinese, Simplified Chinese, and Spanish. Smarter handling of long pasted text Pasting a long block of text no longer clutters the composer — it collapses into a small "Pasted" chip that you can tap to preview the full text, and very large pastes are turned into a file attachment automatically. GitHub Copilot as a provider You can now sign in with GitHub Copilot as a model provider, with a clear consent step before it's enabled. All-day calendar events You can now create, edit, and toggle all-day events through calendar tools. New models Added Claude Opus 5.5, Claude Fable 5.1, GPT-6 Sol, GPT-6 Luna, GPT-6 Astra, and the gpt-image-2.5 image model variants. Bug Fixes - Fixed several causes of conversations getting stuck with "Context is full" or getting mis-compacted, including cases where the app's own size estimate for a model drifted from reality - Fixed the app sometimes losing recent messages or applying sync updates out of order when iCloud sync ran across multiple devices - Fixed backups taking longer than necessary on large accounts by compressing and restoring in parallel; also fixed a slow-but-working WebDAV destination being treated as a failure, and restore falsely reporting items as failed when they simply weren't in the backup - Fixed several crashes: rapid taps on draggable buttons, oversized message previews, AlarmKit on unsupported platforms, and deleting a focused environment-variable row - Fixed provider and model issues: model list not refreshing after OAuth sign-in, switching models mid-conversation leaving the old one retrying, thinking text lost after restart, provider settings landing on a blank screen, and slow, poorly ranked model search - Fixed voice issues: wrong read-aloud voice used, microphone showing a generic error instead of "interrupted by a call," voice input not restarting after a call, and unusable options in some providers' voice model lists - Fixed reply formatting: inline code losing its style or misaligning, equation numbers overlapping formulas, and scrollbars on code blocks that already fit on screen - Fixed all-day calendar events losing their all-day status when edited, and not being creatable as all-day from certain flows - Fixed the storage usage shown in-app not matching the number shown in iOS Settings - Fixed the PDF preview sometimes not being dismissible - Fixed sub-agent conversations occasionally missing a title or not being fully removed with their parent conversation - Fixed the conversation list occasionally showing duplicate or misordered entries, and loading slowly on accounts with many conversations - Fixed a mistap on the "…" menu jumping the conversation to its oldest message - Fixed the app being able to jump to a broken screen on iOS 16 from a Shortcuts action - Fixed an error notice staying on screen after a retried message actually succeeded - Fixed an attachment sometimes staying stuck loading and never displaying

Minis (iOS) 1.13 New: Backup & Restore (Settings → Storage) Export your chats, provider configs, credentials, environment variables, custom thinking rules and shared files into an encrypted .minisbak package, and send it to Files, a network folder, or a remote server (S3, WebDAV, SFTP, SMB, FTP, etc). Restore opens from your configured destinations, lets you browse folders, shows download size and progress, and lets you cancel a slow package open or download. Name your device so backup packages are easier to identify and sort, and delete individual remote backups. Restore merges everything back safely, with automatic rollback if anything goes wrong. New: Personalize your assistant Choose an emoji or image as your assistant's identity icon. The icon appears consistently across chats, and tapping the assistant header opens Soul settings directly. New: Spanish language support Minis is now available in Spanish. Language changes apply throughout the app without requiring a restart. New: Improved accessibility Chat controls are labeled for VoiceOver, and Minis announces when an assistant response finishes. Also new - Third-party OpenAI/Anthropic-compatible providers can be added without an API key - The thinking toggle shows an icon with an "Off" label when disabled - Weather results include precipitation amount, wind gust and direction - Long-press a chat message for "Delete From Here" - Swipe left on a provider, model group, or backup destination to Edit or Delete - Transient provider errors now auto-retry with backoff - The About screen shows the build date next to the version number Bug Fixes - Fixed a file-permission check that could be bypassed under certain paths - Fixed crashes from malformed or mistyped custom provider Base URLs - Fixed math formulas being invisible in dark mode - Fixed shared-scope web apps rendering a black screen - Fixed conversations getting stuck after context compaction or an unanswered message - Fixed a rare crash from an out-of-range message index in long conversations - Fixed images and video thumbnails sometimes appearing squished on first load - Fixed the tool-call status losing its model-supplied title - Fixed the @-mention file list going stale after mounts or files changed - Fixed "Add to input" silently overwriting existing draft text - Fixed stale stream-stall errors, and mid-stream stalls not retrying automatically - Fixed the code-block copy button not responding on iOS 17 - Fixed the masked API key field not being tappable for editing - Fixed Gemini 3.7 Flash and Claude 5 rejecting the thinking-off setting - Fixed the browser automation tool hanging or returning unusable values - Fixed a background tool service exiting while still in use - Fixed shared content being lost or delayed when the target session wasn't active - Fixed paused-session badges being re-stamped or showing the wrong date - Fixed provider sync not repairing iCloud data divergence - Fixed a rare crash from re-laying-out a large Chinese/CJK document preview in the background - Fixed the session list showing a raw tool name instead of a readable summary - Fixed memory-heavy requests sometimes failing unexpectedly during long conversations - Fixed a broken local environment getting stuck instead of recovering automatically - Fixed long inline code in Markdown messages wrapping or highlighting incorrectly - Fixed audio files with no usable audio track failing after being sent to transcription - Fixed backup restore path handling and category counts in some cases - Fixed token usage sometimes being attributed to the wrong model - Improved performance of the provider list and long-conversation compaction - Improved stability of the built-in terminal

Session Groups Group related sessions together in the session list, with a custom icon — either manually or by letting AI auto-classify them. Group cards surface member status alerts, and a mini-bar appears while scrolling through an expanded group so you can jump back to its header. Vision Input: image understanding for text-only models Models without native vision (like DeepSeek) can now "see" images. Minis asks a vision-capable model to describe the picture first, tells you which model did it, and reports clearly if that fallback fails. Custom Thinking Rules Providers now have a "Thinking Rules" section on their detail page — add, edit, delete, and reorder how reasoning parameters get sent per model. Rules sync via iCloud. Liquid Glass across the app On iOS 26, the home screen's floating buttons, the message composer, the tool-status bar, and sent messages all now use Liquid Glass instead of a solid background. Older iOS versions are unaffected. Kimi Code account sign-in Added the option to sign in with a Kimi Code account. Richer parameters for Apple tools Reminders now supports location-based alerts (on arrival/departure) and recurring rules. Notifications support repeating schedules and interactive action buttons. Weather adds minute-by-minute precipitation for the next hour. Reliability improvements Fixed a stuck iCloud sync migration that could show fake progress or never finish. Fixed thinking models stopping mid-stream during long responses. Fixed background tasks occasionally being interrupted by the system, and a rare crash when an external storage volume was mounted. Fixed a launch-crash loop so the app recovers instead of repeatedly crashing. Faster, more reliable prompt caching Fixed several issues that were preventing prompt caching from working as expected across multiple providers, and corrected a token-usage counting bug that could misreport cache hit rate and context usage. Bug Fixes - Fixes long text getting clipped or unscrollable when pasted into the composer - Fixes voice input/output issues, including muted read-aloud not resuming and dropped opening words in recordings - Fixes math formulas with superscripts and subscripts sometimes rendering as plain characters - Fixes provider names occasionally showing incorrectly on a model row - Fixes message bubble heights sometimes being measured inaccurately for image attachments - Fixes HealthKit queries sometimes silently reporting "no data" instead of the real error - Fixes the model list order and refreshes the Codex OAuth model list - Fixes Shortcuts automation notifications (duplicate alerts, missing prompts, permission prompts during a run) - Fixes "Move to…" sometimes losing shared content when the target session wasn't ready yet - Fixes a custom OpenAI-compatible provider rejecting requests over an unsupported field - Fixes several chat rendering issues, including blockquote nesting near images and stale layout heights in the footer area

New: "Ask Minis" via Siri — say "Hey Siri, Ask Minis" to jump straight into a live conversation. New: Kimi Code sign-in support, plus full OAuth login for MCP servers (with iCloud config sync across devices). New: Codex Fast Mode toggle for OpenAI OAuth models. New: Live Activity (privacy mode) now shows real task progress — position, status, and run time. New: Select any text and tap "Read Aloud" to hear it via Minis TTS. Voice correction learning now uses real conversation context, further improving correction accuracy. New: Max output length raised from 64K to 128K tokens. In-browser downloads got a full UX overhaul with floating progress cards and a downloads panel. Stability: Fixed several stability issues in message streaming, Live Activity restore on iPad, HealthKit data reads, and config loading after a device restart. Reduced stutter during long "thinking" streams and scrolling. Bug fixes: - Markdown tables keep their scroll position while streaming - Fixed duplicate thinking bubbles on retry - Fixed invisible scroll buttons in light mode - Fixed navbar title staying blurred after Face ID unlock - Session list timestamps are properly localized - Fixed in-app browser fetch failures - Fixed Mistral chat compatibility issue - Backfilled translations for ~200 previously English-only strings - Added Claude Opus 5 to the built-in model list - Fixed occasional config or chat history issues after a device restart - Fixed occasional unresponsiveness when checking Now Playing info or using speech synthesis - Fixed regenerating a session title not following the latest model order - Fixed some health metric types occasionally being skipped when reading HealthKit data - Fixed onboarding setup steps missing translations in some languages - Reduced app install size

Voice input & output On-device VAD recording with pluggable ASR/TTS providers (Azure, Doubao, Xiaomi MiMo, system online/offline), fail-over across providers, a redesigned voice panel, and an optional AI correction pass using recent conversation context. App Lock with Face ID Lock the entire app with Face ID, Touch ID, or passcode, with a configurable grace period and a privacy screen in the task switcher. model-use passthrough mode Command-line tool calls can now pass raw request bodies, headers, and custom endpoints straight through to a model provider, with warnings surfaced when fields are silently dropped or downgraded. Thinking level in the nav bar A badge to view and change a session's reasoning effort, with Max/Ultra tiers, a picker, and a slash-menu shortcut. Streaming rendering Messages glide into place with word-by-word fade-in; code blocks animate open as they stream. Model Quick Test Test a model's top-3 modalities concurrently, with results shown side by side. Live Activity Play/pause and mute/unmute narration from the Dynamic Island and Lock Screen; fixes stuck timers and stale state after a session finishes. Compact Messages A "Compact Messages" menu entry and a one-tap auto-compact option on the capacity warning. Also in this update - Long-press the new chat button to start a session with a specific model group - Azure OpenAI mode for OpenAI/Responses providers - Modality icons and priority tags on Model Groups - Editable Context Window for any model - Scroll-to-top button in long chats - Print support for web pages, images, Markdown, and text previews - Copy table as image - Session titles summarize both the first and last turns - Cache hit rate shown in the Token Usage sheet - Claude Sonnet 5 and GPT-5.6 (sol/terra/luna) added; Grok list synced - Multiple fixes to the built-in terminal, improving stability Bug Fixes - Fixes the "..." menu refreshing or resetting mid-streaming - Fixes context-window usage showing stale numbers after editing a model or model group - Fixes English text wrapping mid-word instead of at word boundaries - Fixes a bottom gap in the voice panel after edit-and-send - Fixes a 3-second UI freeze when switching from voice to text input - Fixes voice-template model sync deleting API-fetched audio models on launch - Fixes three app crashes caused by dangling view references - Fixes Markdown code block spacing inconsistencies - Fixes voice input panel gestures conflicting with transcript scrolling and editing - Fixes the chat input bar retaining stale height across re-entry - Fixes terminal crashing after a rootfs reset - Fixes silent playback in the System TTS/ASR path - Fixes thinking not disabled for adaptive-reasoning models set to "off" - Fixes Gemini models failing when a response contains an empty text segment during tool calls - Fixes reasoning mode not working for some third-party models - Fixes retried messages occasionally disappearing from chat history - Fixes Token Usage counting inaccuracies on reloaded sessions - Fixes a crash on iOS 26.5 caused by a removed WebView API - Fixes table columns collapsing or clipping emoji during streaming - Fixes Bluetooth keyboard send leaving the software keyboard on screen - Fixes a memory leak in iCloud sync's per-session cache - Fixes idle CPU sitting at 100% in the session list - Fixes a slow cold start when opening a session - Fixes main-thread hangs during long-message streaming

What's New 1. MCP Integrations — Connect Any Tool to Your AI Minis now supports the Model Context Protocol (MCP), letting you connect external tools and services directly into the agent loop. Add MCP servers via Settings → MCP Integrations, import configs from JSON, or manage them from the shell with the new  minis-mcp-cli  command. 2. SOUL Personality Customize your AI's personality — name, style, language, and character — directly in Settings. Everything is saved to SOUL.md and persists across sessions. The sidebar title, default session name, and typing indicator all reflect your Soul's name. 3. iCloud Sync V2 Upgraded to record-level sync. Providers, Model Groups, and environment variables now sync independently per entry. Fixes issues where deleted configs were resurrected by iCloud, local API keys were overwritten on merge, and active sessions had their content overwritten mid-run. 4. Providers OpenAI: Generate images directly in chat using gpt-image-2 via OpenAI OAuth / Codex instances. New xAI provider with one-tap SuperGrok / X Premium+ login. 5. Memory Toggle A new global memory toggle in Settings controls whether new sessions enable memory by default. Turning it off fully suppresses memory injection and hides memory tools. 6. Tool Calls The agent loop now dispatches multiple tool calls in parallel, making multi-step tasks significantly faster. Long-press any tool capsule to re-run from that exact point, or copy the tool call details. Bug Fixes • Fixed multiple chat crashes including session switch UAF, notification deeplink wake, and iPad sidebar collapse jank. • Fixed generated content being lost after Stop or Retry. • Fixed vision models not seeing images returned by read_image or browser_use. • Fixed concurrent browser_use calls trampling the same WebView across tabs. • Fixed Codex OAuth non-streaming calls triggering HTTP 400 Stream must be set to true. • Fixed orphaned tool_use and tool_result blocks being sent in Anthropic requests. • Fixed reasoning models returning empty text when generating session titles. • Fixed slash menu disappearing in over-input mode and installed Skills not appearing in the list. • Fixed token counts being inflated by providers that report cumulative usage on every SSE chunk. • Fixed compact running in an infinite loop. • Fixed image generation requests timing out at around 120 seconds. • Fixed camera captures being stored as oversized lossless PNGs.

New Features 1. minis-config CLI: agent edits app settings from terminal with audit and revert. 2. Environment variable privacy mode: env values masked in shell output before reaching model. 3. Full minis:// deeplinks: external links jump into Settings and file browser. 4. New Chat onboarding: workspace folders and file browser entry on start. 5. File mention ranking: @ mention ranks closer to real usage. Bug Fixes 1. Empty responses: GPT-5.5 blank replies now show error with retry. 2. Reasoning fallback: fixed early stop from fabricated placeholders across providers. 3. Long-session stability: 200-turn cap with resume; cross-provider tool 400 fixed. 4. DeepSeek V4 compatibility: fixed misflagging as vision-capable. 5. Image attachments: picker/drop images keep original bytes, no JPEG. 6. Streaming scroll jumps: fixed half-screen jump and broken table links. 7. Image preview transition: full-screen image no longer freezes mid-animation. 8. Misc polish: empty state visuals, browser preview sizing, settings link routing. Performance 1. Tool streaming heat: CPU and device heat drop on large tool inputs. 2. Long-chat smoothness: smoother scroll and streaming; logs shrink dramatically.

New Features 1. Added support for DeepSeek V4 with thinking toggle. 2. Added support for GPT-5.5. 3. Added provider config export and import. 4. Added tool loop detection to prevent Agent infinite loops. 5. Added drag-to-reorder for models, providers, and groups. 6. Per-instance image generation endpoint mode (auto / images / chat). 7. Clickable URLs in terminal with link preview support. 8. Confirm prompt before overwriting environment variables. 9. Confirm prompt before deleting a model entry. Bug Fixes 1. Fixed iCloud sync reliability across multiple scenarios. 2. Fixed image attachments being truncated during history compression. 3. Fixed Claude 4.6+ adaptive thinking and request headers. 4. Fixed browser_use deadlock and missing preview.

New Features 1. Added full Traditional Chinese localization. 2. Support custom web viewport with per-session override and UA-mismatch warning. 3. Enhanced chat input with @-mention file picker, Insert Line Break menu, Return-key behavior toggle, and keep-screen-awake setting. 4. Added paged image gallery with drag-to-dismiss and pinch zoom. 5. Support Apple Photos import and save to specific album. 6. Support OpenAI image generation and editing. 7. Optimized Skill system to sort by priority and usage frequency. 8. Added task-completion app icon badges. 9. Added email feedback entry in Settings. 10. Support double-tap and right-click to copy Markdown table cells. 11. Tool sheet terminal button now prefills the current shell command. Bug Fixes 1. Fixed streamed text tail characters and bold formatting going missing between tool calls. 2. Fixed math-heavy markdown previews freezing the app on open. 3. Improved iSH terminal stability across iOS 16 and iOS 26 — fixed blackouts, second-open hang, paste, keyboard focus, and memory leaks. 4. Fixed Dynamic Island and keep-screen-awake not releasing after tasks finish. 5. Fixed Stop button not killing long-running processes like sleep. 6. Improved iCloud sync reliability — fixed empty-session uploads, cross-device deletes, custom model preservation, and drag-reorder races. 7. Fixed File Browser not refreshing the Files app after copy/move/delete. 8. Fixed shell_execute swallowing stderr, browser_use ignoring UA settings, and /compact missing the last turn. 9. Fixed tool execution status not persisting across app restart. 10. Fixed Report Content emailing the wrong address. Performance 1. Optimized math formula rendering with lazy loading — markdown previews are now scrollable instantly. 2. Optimized terminal rendering with incremental updates and batched TTY output. 3. Moved streamed markdown re-parse off the main thread. 4. Optimized iCloud sync to run off the main thread with per-session etag caching.

New Features 1. Mount external folders: Mount iCloud, Files-app, and other external directories into the Minis workspace that Agent can help you reading/editing. 2. In-app resource previews: The agent can use minis-open to open chat images, Markdown/HTML, and media directly in the right in-app preview. 3. Apple Health improvements: New delete subcommand removes entries created by Minis, add more types support. 4. Cross sessions tool: Minis can read or search your history sessions by embed tool. Bugfix 1. Fixed network tools hang in some cases. 2. Fixed terminal can not interactive on sime iOS versions. 3. Other crashes and performance improvements.

Terminal Overhaul - Full CJK and emoji input support with pixel-perfect grid alignment - Complete external keyboard support: Ctrl/Alt combos, arrow key modifiers, text selection - Fixed terminal becoming unresponsive after backgrounding; added command history Model Groups - Session defaults: set reasoning level and context window limit per group - All model errors shown when group fallback is exhausted - New "Always Fallback" strategy — switch immediately on any error Drag & Drop - Drop files and images directly into chat from Files or Photos iCloud Sync - Major performance improvements — sync deferred during AI processing - Fixed multi-device edit conflicts and data loss Third-Party Coding Plan Support - Manual OAuth token entry for MiniMax, Kimi, and other compatible services - Fixed token loss during provider import/export Safari-Style Web Preview - Live tool streaming and inline web previews Files App Integration - Access Memory, Skills, and shared folders directly in the system Files app Bug Fixes & Polish - Smoother streaming text rendering — fixed flickering and overflow - iPad layout and dark mode improvements - Fixed input field not clearing after CJK IME confirmation

1. Smarter context management — Tiered offload/compact/exhausted policy adapts to each model's actual context window instead of applying a fixed threshold. 2. Calendar & Reminders write support — You can now update, delete, and complete existing events and tasks, not just create them. 3. Input improvements — Drag to reorder attachments before sending; long pasted text auto-attaches as a file. 4. OpenAI Responses API merged — Now an API Format option inside the OpenAI provider, no separate entry needed. 5. Stability & performance — Fixed multiple crashes (URLSession, SwiftUI cell height, TableAttachment), restored thinking dots and sheet animations, faster Terminal startup, and more reliable speech input.

- New /thinking command with Off / Low / Med / High intensity levels. - NFC tag read/write reliability improvements - Model list automatically enriched with modality and capability metadata - Fixed spurious session reload triggered by iCloud sync - apple-vision tool gains similarity and overlap subcommands - Other bugfix

- iCloud Sync — Sessions, files, and provider configurations now sync seamlessly across all your devices. API keys sync securely via Keychain. - Session Pinning — Long-press any session to pin it to the top of your list. - Launch Session Setting — Choose how Minis starts: auto-resume, last session, or always open a new chat. - Shared Directory — New  /var/minis/shared/  folder persists files across sessions and is accessible from the Files app via a FileProvider extension. - Smoother Scrolling — Rebuilt the message list layout engine from scratch to eliminate scroll jitter and improve rendering stability. - Bug Fixes — Fixed compact divider incorrectly graying out new messages, multiple crash points (watchdog kill, Range bounds, CollectionView reuse), and Markdown table height calculation.