# Skales - AI Agent & Tasks — Automation with any AI model > Local-first AI assistant for private, tool-integrated conversations and tasks. (iPhone/iPad app by Mario Simic.) - Source: https://appshunter.io/ios/app/skales-ai-agent-and-tasks/id6763328966 (this page in markdown: same URL + `.md`) - Developer: [Mario Simic](https://appshunter.io/developer/1895266452) - Category: Productivity, Utilities - Price: Free - Age rating: 4+ - Requires: iOS 15.1 · 156 MB - Languages: American English - Released: 2026-07-09 - Data updated: 2026-09-01 - Monetization: free - User reviews in markdown: https://appshunter.io/ios/app/skales-ai-agent-and-tasks/id6763328966/reviews.md ## What is Skales? Skales is an AI agent that does the work, not just the talking. Give it a job and it plans the steps, uses real tools - calendar, tasks, web, files, contacts - and delivers a finished result, asking your permission at every step. It runs on the AI of your choice, pairs with your computer, and keeps everything it knows about you on your device. No terminal, no setup. Install in seconds. WHAT AN AGENT DOES THAT A CHATBOT CANNOT "Plan my Friday" becomes a structured schedule in your real calendar. "Research this and write me a briefing" becomes cited sources and a document. "Open a ticket and tell me what they answer" becomes an API call, made with your approval, with the reply shown to you. Turn on Pocket Run and the agent keeps working across many steps until the job is done - it does not stop at a wall of text. EVERY MAJOR AI, ONE APP Start free with Skales IQ - just your email, no API key, no card. Or bring your own key for OpenAI, Anthropic, Google, Mistral, OpenRouter and more. Chat with Claude, GPT, Gemini and dozens of others, pick the reasoning depth per conversation, attach photos and files, and switch models anytime without losing your session. AUTOMATION WITH A HARNESS Skales is built as an agent harness: tools, memory, skills and permissions working together. Built-in agents (planner, coder, studio) plus custom agents you build yourself - or a whole team of agents from a preset. A curated skill library ships with the app, from disciplined bug diagnosis to research with cited sources, and the agent reaches for the right one on its own. Every action that touches the outside world asks first. Connect Home Assistant, Spotify or WordPress with your own account. TALK TO IT - IRIS ORBIT A full voice mode: talk to Iris Orbit or to your Buddy and it talks back, hands-free, with your tools and approvals working exactly as they do in chat. MAKE THINGS, NOT JUST ANSWERS Flow turns a brief into a finished document, deck or motion piece: 74 design aesthetics, your own brand kit, export as PDF, image or self-contained HTML. Lio builds a clickable app or prototype from a sentence - and shows you the plan before it starts. Studio generates images and video and designs social posts you can share to the community feed. A COMPANION YOU CAN READ Turn on the companion and a slim mood bar lives in your chat: warm when the work goes well, cool when it does not. Tap it and Conscious tells you what moved it today - no notifications, no sounds, it just shows. A MEMORY THAT GROWS Skales remembers what matters across conversations and adapts to you - stored on your device, fully under your control. YOUR COMPUTER, FROM YOUR POCKET Pair with the free Skales desktop app for Windows, macOS and Linux over your WiFi, and your phone reaches your computer's AI, sessions, files and tools. Or open the in-app web interface from any browser on the same network. PRIVATE BY ARCHITECTURE, NOT BY PROMISE Your conversations and memory stay on your device by default. With your own key, requests go straight from your phone to the provider you chose - Skales runs no servers in between. No account required. No ads, no tracking, no data sold. Optional online features are clearly marked and off by default. And when you want zero cloud: Skales Local (beta) downloads an openly licensed model once and answers directly on your iPhone or iPad - no key, no connection, even in airplane mode. Each model card says plainly what it can and cannot do. DISCOVER A live community feed, topic Spaces, and a personal news briefing pulled straight to your phone. DESIGNED FOR EVERYONE Grandma approved AI agent. No technical knowledge needed. 12 languages. Need help? Email request@skales.app - more at skales.app. ## Key features - Local-first AI agent - Supports multiple AI models - Voice mode for hands-free interaction - Integrates with real-world tools - Image and video generation - Growing memory across conversations - Community feed and topic spaces ## Recent user reviews (3 of 3) All indexed reviews: https://appshunter.io/ios/app/skales-ai-agent-and-tasks/id6763328966/reviews.md ### 2/5 — Constant Crashes *2026-08-16* I downloaded the desktop version and I was pretty impressed. I then proceeded to download and setup the IOS 26 version and, unfortunately, I’m met with constant crashes and can’t even get a prompt in or a response from the app. I don’t know what I’m doing wrong. I have ai providers linked and everything seems fine on paper but it doesn’t work regardless of the provider i choose. It’s unfortunate because I do plan to use it to the fullest capabilities and now my customer satisfaction has been low. **Developer response:** Update: the crash you hit is fixed. 2.9.15 addressed the iOS 26 case. If it runs clean for you now, I would appreciate you revisiting your rating. If you still see problems with any provider, Settings > Advanced > Logs and Diagnostics captures it and request@skales.app reaches me directly. ### 5/5 — 5+ *2026-08-09* Hi, good luck! I’ve started learning this kind of software with your product — it looks intuitive and simple, and all the necessary information and instructions are available. I’m completely satisfied, thank you for your work! **Developer response:** Thank you, that really means a lot as a solo developer! Skales Mobile started as a remote control for the Skales Desktop app. Based on what users kept asking for, we are turning it step by step into a proper standalone app, as far as the operating system allows. More local models on device are part of that path, and more is coming with every release. (You can submit your feedback in Settings > Help > Report a Bug.) Glad the setup and the docs worked for you. If anything ever feels unclear, write to request@skales.app. ### 5/5 — Incredible Feature Set! *2026-07-24* Wow! One of the best and most full-featured Agentic AI apps I’ve ever used! Integrations for mail, Notion, Home Assistant, Todoist, Spotify, Google Calendar, YouTube, Wordpress, and more. Great UI. Local Gemma models in beta or BYOK with Openrouter (including TTS!), Mistral, Ollama, and basically any you want, and you can set up different models and providers for different roles. Agent teams, MCP, and skills etc all available. On device file reading and writing, API-free weather, reminders, device info, contacts, so much more. Extremely impressed and excited to see how this app improves!!! Would love to use custom TTS voices eg ones running on my Mac (or even on device like Kokoro or Kitten etc) in the future but that’s really the only thing I can think of that’s not already included. Fantastic agent harness and frontend in one! **Developer response:** Thank you, this genuinely made my week. Skales is a solo project, so a review with this much detail in it means a lot for me. Two things are already on the way. NVIDIA is coming as a provider, which opens up a lot of strong open models. And a background runner is planned, so your agent keeps working when you close the display instead of dying with it. Custom TTS voices are noted (you will get it!), Duplex Voice Chat, and local or on-device options are exactly the direction I want to take that. Thank you for naming it, that is the kind of feedback that actually shapes the roadmap. Beyond those, my focus stays on more features, better stability, and a smoother experience throughout. Thanks again for taking the time to write all of this, I appreciate it a lot! If not already done, you can connect the app to Skales Desktop for a big boost in tool use inside chat, and so runs keep going while the phone is off. The desktop app is signed and notarized on macOS, and it ships with a migration importer. ## Frequently asked questions about Skales ### Does Skales have ads? Skales is free to start and does not contain advertisements. Its monetization model is free with optional in-app purchases for advanced features or model access. ### What AI models can I use with Skales? Skales supports a wide range of AI models including Claude, GPT, Gemini, Mistral, and dozens more. You can pick your preferred model and reasoning depth for each conversation. ### Can Skales run offline? Yes, Skales Local allows you to download an openly licensed model and run it directly on your iPhone or iPad. This feature works in airplane mode, meaning no internet connection is required for basic chat functionality. ### How does Skales handle privacy? Skales is designed with local-first privacy in mind. Your conversations stay on your device, and optional cloud features are clearly marked and off by default. No account is required to use the basic features. ### What kind of real-world tasks can Skales perform? Skales can interact with tools like your calendar, tasks, web search, files, and location, always asking for your permission first. It can also act on the web by calling APIs or sending forms, and includes built-in agents for planning and coding. ### Is Skales suitable for children? Yes, Skales is designed for everyone, including children. It has an age rating of 4+ and requires no technical knowledge to use, making it accessible for users of all ages. ### How often is Skales updated? The latest version of Skales is 2.5.5, and it was last updated on July 24, 2026. Updates typically focus on improving features, adding new capabilities, and enhancing performance. ### What devices does Skales support? Skales is available for iPhone and iPad. It is designed to run seamlessly on these Apple devices, offering a consistent experience across platforms. ## Version history (last 5 releases) ### 2.9.15 — 2026-08-30 This release fixes a crash that could stop Skales from opening at all on iOS 26. Also fixed: A long answer is now read to the end. The safety timer that watches read-aloud used to cap every reading at two minutes, so anything longer than about 1,300 characters was cut off mid sentence. The timer now watches for a reading that has actually stalled, so length is never punished and a genuine hang still recovers. Every agent in the chat picker is reachable. The list under the height cap could grow past the screen with no way to scroll, and "Manage" was the last row, so the way out disappeared with it. The rows scroll now, the title stays put, and Manage is a fixed footer. The goal sheets in Chat and Code had the same flaw and got the same fix. A start that fails names itself. The app now carries a safety net above every screen, the preference store survives a failed start with a named fallback instead of dying silently, and the entry decision can no longer leave the app on a black screen. The last start's steps travel with the crash report, so the next mystery is a measured one. Small on device models are no longer handed instructions they cannot follow. Very small local models were sent a protocol built for larger ones and could repeat themselves until stopped by hand. Models that small no longer receive it, and a reply that starts looping stops with a reason on screen. Larger models are unchanged. Typing in Flow stays visible. The model, style and video pickers now move out from under the keyboard, the way every other input surface in the app does. Read aloud follows the language of the text, not the language of the interface, and a voice you chose yourself always speaks, even when the text is in another language, because that was your choice and not an accident. New: a female Spanish voice on the device. Sharvard joins the on device voices, downloadable from the same list as the others. ### 2.9.10 — 2026-08-29 Teams. A new Teams screen holds small shared rooms with other people, phone to phone over the Skales relay. Create a room, pass its code or QR while the door is open, and confirm every newcomer by name and key words. The member sheet keeps who belongs and who is here now apart, your own messages carry a quiet delivered to X of Y, a message that could not leave the phone shows Failed with tap to retry, and leaving a room deletes it here. Your name and picture live on this phone. The desktop app draws the same rooms. One of your own agents can sit in a room as a member. Writing @AgentName sends the message and a signed assignment, only the owner's device ever runs it, and you decide per agent who may call it. Every device shows the same task card through its states: waiting for approval, running, the answer, or a named refusal. Every room has one shared board. Rows go to people or agents, and nothing runs until every member has approved exactly that list. Your own row can be ticked off or taken into a real chat seeded with the board and the recent thread. When two people edit at once the newer version wins, and the other person sees it instead of losing work silently. Studio can start a 3D model from a picture, not only from a description, and it offers only the models that can really begin from what you chose. Every refusal names which provider said it and why, in your language. Plugins can come over from your paired computer instead of only from a backup, with the same permission lines you see after installing. A plugin arrives switched off. A plugin page may call the tools its own manifest lists and puts up the same approval card a chat turn does. It still has no network and can never call a model. Ask for six beaches or five recipes and the answer keeps its prose, with a row of cards underneath: preview picture, name, one sentence, tags and the link. Place cards open in Apple Maps. Discover posts carry a Translate button that turns text, link title and poll into your language, on the model you already configured. A Flow film can go to Discover from the export sheet. Settings gained an Add-ons page that decides which of the nine optional entries the menu offers. Approval cards now say what is about to happen in a sentence instead of a function name, in all twelve languages, with the raw call folded away under Details. Conscious can say what it is thinking about and drifts with the shape of the day, on this phone. Fixed: a spoken conversation now pauses when you leave it instead of reading an answer into your pocket, and an incoming call that takes the microphone is named on screen. A paused goal really stops. A daily task stays daily. Editing a task no longer undoes what happened while you were looking at it. A design that stopped early says which of four reasons applies. Providers explain a refused image in words. Every switch in Settings tells a screen reader which setting it is. Reminders, approval cards and eight languages are written with their own letters again. ### 2.9.0 — 2026-08-26 Fable The first launch now asks who you are before it asks for keys. New to AI, or at home in it: the answer shapes everything after it. Newcomers get plain words and a guided provider choice, experienced users get dense screens and can skip whole blocks. A phone already paired to a desktop is recognised and told that the computer does the thinking. Setup settles the everyday. Look, light or dark, an accent tone, your city, a voice you can hear before deciding, and notifications explained before the system asks. The whole run works offline, the back button behaves on every step, and killing the app resumes exactly where you stopped. It ends on three cards showing what Skales is actually for. New in this release: - Guide. One page with "I want to..." paths and a catalogue of what this phone can do, each line showing live whether it is ready, with example prompts that land editable in the chat box. - Templates. The desktop catalogue, bundled and offline. - Flat. A second look beside the house one: grey and white, one accent colour, a text-only menu. - Memory search that ranks instead of filtering. Two words in any order find the entry that carries both. - Setup can be replayed later. Answers arrive prefilled, and finishing changes only what you edited. - Your experience level travels with the pairing. Answer on either device and the other one learns it. Fixed: - A profile field you clear stays cleared, on this phone and on the paired desktop. - Short searches find things again. "C#" matches, and a single Japanese or Chinese character filters the list. - When the desktop access token was renewed, the screen showed a refusal and no way back. It now puts the address field and QR scanner back up, and a key button in the header leads there at any time. - A slow image gets an honest answer instead of hanging, and a retried message brings its pictures back into the composer. Six pollers used to keep fetching behind the home screen. They all pause now and catch up the moment you return. ### 2.8.4 — 2026-08-25 The reasoning view is repaired: a thinking model's trace stays under the answer through every step of a task and is still there when the answer is finished - it no longer flickers back to "Reasoning" while the agent works. Sessions in History no longer show dates from the future. Small fixes throughout. ### 2.8.3 — 2026-08-20 Skales 2.8.3 is a repair release, and most of it is about things arriving whole. A thinking model's reasoning is back under the answer. The fold only ever filled for models that write their thinking into the reply itself, which on this phone meant the ones running on the device. Everything in the cloud sends that trace on a channel of its own, and the phone was throwing the channel away without reading it. It reads it now, asks for it where it has to be asked for, and the fold is still there when you come back to the conversation after a restart. An answer stops saying everything twice. A reply that used a tool streamed its second round on top of the first, so the text stood there doubled until the turn ended and snapped back into place. Every round now replaces what came before it, in the chat, in the browser window and in Flow. A long thinking answer also stops getting sticky: the screen was re-reading the whole buffer for every single token, which on a long answer running on your own phone was more work per token than the model itself was doing. A file Skales writes is a file you can open. Asking for something as a document reached a tool that did not exist on the phone, so the answer came back as paragraphs in the scrollback. There is a document tool now, and pictures, clips, voice files and visuals leave a real file behind as well, each with a card under the answer. The card survives leaving the chat mid answer, and a turn that fails after the file was written still shows it. Over the bridge to a paired computer, a document written on that computer used to arrive as a line of text; it comes as a card that says where it lives, and tapping it fetches the file and opens it. A refusal says what it was. A provider that turned a request down mid stream produced an empty answer and a line saying the request was rejected, with the actual reason discarded on the way. The reason is kept now: a model that does not exist says so, a rate limit says so and offers to try again, and the full response is on the diagnostics page. That page can be read as well. Every line used to be cut off at the width of the screen with the useful part far off to the right, entries sat out of order and some carried a date in the future. Lines open, wrap and can be selected, errors are open on arrival, and the order is the order things happened. A session in History says who answered it. The row carried a date without a time and never named the provider or the model. It now reads like the desktop's, with time, provider, model and count; older sessions leave out what they never recorded. And a portrait picture stays portrait in Discover, because the shape now travels with the picture instead of being guessed at as a square. ## Related topics [ai models](https://appshunter.io/ios/topics/ai-models) · [top ai tools](https://appshunter.io/ios/topics/top-ai-tools) · [cerebrocore](https://appshunter.io/ios/topics/cerebrocore) · [cappy ai](https://appshunter.io/ios/topics/cappy-ai) · [ai peeps](https://appshunter.io/ios/topics/ai-peeps) · [cucle ai](https://appshunter.io/ios/topics/cucle-ai) · [circadian ai](https://appshunter.io/ios/topics/circadian-ai) · [ai programs](https://appshunter.io/ios/topics/ai-programs) · [dole ai](https://appshunter.io/ios/topics/dole-ai) · [local ai studio](https://appshunter.io/ios/topics/local-ai-studio) · [volter ai](https://appshunter.io/ios/topics/volter-ai) · [ai question answer](https://appshunter.io/ios/topics/ai-question-answer) · [niche ai](https://appshunter.io/ios/topics/niche-ai) --- *Data collected daily from the US App Store and indexed by [AppsHunter](https://appshunter.io/). User reviews are verbatim App Store reviews. Ratings, prices and chart positions refresh continuously; this snapshot is from 2026-09-01.*