Privacy-First AI for Professionals
Run AI completely offline, connect to your own servers, or use cloud providers. All in one professional app.
Advanced AI
Inspired by Andrej Karpathy's llm-council project, Privacy AI now supports the LLM Council concept natively. Create a council of AI models, such as GLM, Minimax, Kimi, and even local models on-device. You can have all of them respond to the same question simultaneously. Because Privacy AI already connects to multiple AI providers and models at once, it is uniquely suited for this: no external orchestration layer is needed. Use LLM Council for multi-perspective research, structured debate, critical thinking exercises, and business strategy analysis where hearing diverse AI viewpoints leads to richer, more balanced insights.
Cloud & APIs
Privacy AI can now run as a local OpenAI-compatible API server, exposing all your configured models through OpenAI compatible service. External applications that supports the OpenAI API can connect to Privacy AI and use your local GGUF models, MLX models, Apple Intelligence, or remote API providers through standard endpoints: `/v1/chat/completions`, `/v1/models`, and `/health`. The server supports both HTTP and HTTPS (with self-signed certificate), streaming responses via Server-Sent Events (SSE), and includes QR code sharing for easy device connection. This transforms Privacy AI into a universal AI gateway and keep secret of your API Keys. Find it under "Models & Services" in the menu.
Developer Tools
You can now open a full interactive shell terminal from any chat via the Shell option in the chat's context menu (â‹Ż). On iOS, a sandboxed terminal powered by ios_system launches within the chat's workspace directory, giving you access to file and text-processing commands without leaving the app. On macOS, a full interactive zsh session opens with access to all system commands. The terminal displays in full screen with a black background, green prompt, and ANSI color support.
Local Models
Privacy AI now includes an on-device MLX Audio engine powered by MLX-audio, offering models across TTS, STT, VAD, and speech-to-speech categories. Highlights include Qwen3 TTS 0.6B (multi-language, voice design) and Qwen3 ASR 0.6B (multi-language streaming transcription), and more. Download and manage models from the new MLX Audio Models section under Local Models, then select TTS or ASR Audio Settings to use these models for voice output.
Local Models
Privacy AI now supports continuous, long-lived streaming transcription powered entirely by local MLX STT and VAD models. On top of that, you can enable real-time translation so the spoken words are translated into another language on the fly as audio or video plays. Both transcription and translation run fully on-device with no cloud connection required, keeping everything private.
Advanced AI
Privacy AI now supports interleaved thinking (reasoning expansion) for advanced reasoning models from multiple providers, such as DeepSeek (deepseek-reasoner), OpenRouter (minimax-m2, deepseek-v3.2, gpt-5.2, claude-opus-4, kimi-k2-thinking), MiniMax (minimax-m2), Kimi/Moonshot (kimi-k2-thinking), and Google Gemini (gemini-3-pro-preview). This feature preserves the model's internal reasoning process across multi-turn conversations and tool calls, improving reasoning continuity, multi-step problem solving, complex task planning, and decision-making quality. The app automatically detects and handles provider-specific reasoning formats (reasoning_content, reasoning, reasoning_details) and provides helpful error messages when models require this setting to be enabled. Enable Interleaved Thinking in chat settingsx for immediate use, or in model settings for persistent configuration across all future chats.
Document & Media
The Reader now includes a built-in chat interface that lets you have AI-powered conversations directly with your documents. Switch between "Read" and "Talk" tabs using a segmented picker at the top of the screen. In Talk mode, select any AI model and ask questions—the full document content is automatically linked to your chat session, giving the AI complete context to provide accurate, document-aware answers. Perfect for analyzing long reports, researching articles, or extracting insights from technical documentation without leaving the reading experience.
Advanced AI
Introduces the long-awaited *Memory* feature, allowing you to add, edit, and delete memories that persist across all models. Switching between models no longer resets your personal context—your AI now truly remembers you. All memory data are stored securely on-device and optionally synced via iCloud for seamless continuity. Full privacy, full control.
A complete professional AI toolkit, all on-device.
LLM Council now includes a Game Theory mode where council members engage in strategic negotiation across multiple rounds. Each member reasons about the positions of others and adjusts their stance before the chairman evaluates the final outcomes. This produces more nuanced conclusions on complex or contested topics by surfacing trade-offs that straightforward deliberation can miss.
Claude 3.7 and Claude 4+ models now support native interleaved thinking during multi-turn tool call conversations. When enabled via the existing "Include Thinking in History" setting, Privacy AI preserves Claude's internal reasoning blocks (with signature) across tool execution chains using the `interleaved-thinking-2025-05-14` beta header. This completes comprehensive interleaved thinking support across all major AI providers: DeepSeek, Gemini, Kimi, OpenRouter, MiniMax, and Claude—ensuring consistent reasoning behavior regardless of which API you use.
AI responses that contain Mermaid code blocks now render as native charts and diagrams directly in the chat, with no WebView required. Flowcharts, sequence diagrams, Gantt charts, and other Mermaid diagram types are all supported. A toggle button on each diagram lets you switch between the rendered chart and the raw source code at any time. Diagrams automatically re-render when you switch between light and dark mode.
All protocols: OpenAI, Claude API, HuggingFace API, and OpenAI Responses API—now support full network traffic inspection through mitmproxy. This makes it easy to diagnose protocol issues, inspect raw request/response payloads, and understand how data flows through each provider. Enable proxy mode under Settings > Proxy.
The search_news tool now supports five news providers and they all offer free tiers with no cost. 'NewsData.io' provides 200 credits per day. The 'News' API provides approximately 100 requests per day. 'GNews' provides 100 requests per day. 'Currents' API provides 1,000 requests per day. Configure your preferred provider from the search_news tool settings.
The new `manage_appstore_connect` tool lets you manage your App Store Connect account directly through AI chat. You can list apps and builds, read and respond to user reviews, create and update app versions, submit for App Review, manage TestFlight beta groups and testers, browse in-app purchases and subscriptions, create promotional offers and offer codes, set up winback offers, and pull sales and finance reports. Configure the tool with your App Store Connect API credentials (Issuer ID, Key ID, and .p8 private key) from the tool settings.
The new `edit_file` tool lets AI make precise string replacements in workspace files without overwriting the entire file. AI specifies the exact text to find and the replacement text, using surrounding context to guarantee uniqueness when the same string appears multiple times. This is the safe, reliable way for AI to patch scripts, update config values, or fix typos in workspace files - no risk of accidental data loss from broad overwrites.
The new `run_shell` tool lets AI execute shell commands directly inside the workspace directory, unlocking powerful text-processing and automation workflows. On iOS, approximately 150 commands are available, including `ls`, `grep`, `sed`, `awk`, `find`, `tar`, `curl`, and more - with full support for pipes, redirects, and multi-command pipelines. On macOS, AI has access to the full native zsh environment with all system commands. All execution is sandboxed to the chat directory, keeping your files and system safe.
Privacy AI now includes a powerful `grep_file` tool that brings professional-grade text search to workspace files. Search through large log files, data exports, and documents using full regex patterns without memory concerns - the tool streams files line-by-line to efficiently handle files up to 10MB. Advanced features include invert matching (-v) to find non-matching lines, count-only mode (-c) for quick statistics, and separate before/after context (-B/-A) to see surrounding lines. Search multiple files at once using glob patterns like `*.log` or `data-*`, and save results to workspace files for later review. The tool integrates seamlessly with JavaScript REPL via `require('grep')` for scripting workflows. Case-insensitive search by default makes it AI-friendly, while full regex support enables complex pattern matching for log analysis, data filtering, and content discovery workflows.
Privacy AI now supports direct database connections with the new `search_database` tool. Connect to MySQL and PostgreSQL databases to query data, inspect schemas, and analyze information without leaving the app. Features include connection testing, table listing, schema inspection, and read-only SELECT query execution with built-in SQL injection protection. The tool supports complex queries including JOINs, CTEs, and aggregations, with SSL/TLS support for secure connections. Multiple database connections can be saved and managed for quick access.
The `run_javascript` tool now supports executing JavaScript code from workspace files. AI can save large or reusable scripts to the workspace once and execute them multiple times, achieving huge token savings for scripts over 500 lines.
Get comprehensive cryptocurrency analysis with the new 'get_crypto' tool. Access real-time prices for 2000+ cryptocurrencies (BTC, ETH, SOL, and more), along with 10-day historical OHLCV data. The tool provides 11 statistical metrics including volatility, max drawdown, and range analysis, plus technical indicators like SMA10/20, EMA12, RSI14, and trend detection. Multi-API fallback architecture (TwelveData → Tiingo) with built-in rate limiting ensures reliable data delivery.
Get forex data with advanced analytics using the new 'get_exchange_rate' tool. Access real-time exchange rates for 150+ currency pairs, 10-day historical data, and 11 statistical metrics. A unique feature provides correlation analysis for non-USD pairs (e.g., EUR/JPY shows how EUR and JPY move together against USD using Pearson correlation). Multi-API fallback and built-in rate limiting ensure maximum reliability.
It supports new tools to analyze data: 'analyze_data', 'correlate_data', and 'regress_data'. All three tools automatically detect file separators (comma, tab, semicolon, pipe). Numeric parsing is now currency-aware, automatically handling values with currency symbols ($, ¥, €, £), thousand separators (1,234,567), and accounting negative format like (100) to -100. For discrete variable analysis, the value_counts action now provides advanced statistics including Shannon entropy (distribution uniformity), Gini impurity (category homogeneity), and mode percentage, giving you deeper insights into your categorical data.
The new 'create_chart' tool enables AI to generate visual charts and data visualizations during conversations. AI can create charts to illustrate data trends, compare values, or visualize concepts, making complex information easier to understand at a glance.
The web search tool now supports 5 search engines: Serper, Brave Search, LangSearch, Tavily, and Local Search. Configure your preferred engines in Tool Settings with drag-and-drop priority ordering. In Auto mode, Privacy AI automatically tries engines in order until one succeeds, ensuring reliable search results. Each engine can be individually enabled or disabled, with API keys managed directly in settings including one-click "Save and Test" validation. Local Search remains always available as the final fallback.
You can now configure how much context each tool call result can consume. The new `toolResultTokenLimit` setting prevents any single tool from overwhelming the context window—especially important for models with large contexts (80K-128K+). The default limit is automatically calculated as context / 6 (minimum 1000, maximum 24000 tokens), and it intelligently adjusts when you change the model's context size. This keeps your AI responses focused and high-quality by ensuring tool results stay balanced with conversation history.
Privacy AI now includes a smart tool discovery system that enables progressive tool loading instead of sending all tool definitions upfront. The new `Auto Tool` lets AI dynamically discover available local and MCP tools on demand, retrieve detailed documentation, and execute them—reducing context overhead and enabling true agency. This is a major step toward full autonomous agent capability where AI can intelligently explore and utilize its environment.
Three new search tools for books, podcasts, and games. Search books via Google Books with NYTimes bestseller rankings. Discover podcasts through iTunes/PodcastIndex with episode transcripts. Browse Steam store for games, pricing, deals, and news. All tools work without setup and include URL citations for easy reference.
You can now set a maximum number of tool calls per response in each model’s settings. This gives you precise control over the trade-off between thorough reasoning and cost/performance. Higher limits (10–20) allow deeper research flows with multi-step tool use; lower limits (3–6) prioritize faster and more economical responses. This setting is saved per model and preserved inside chat configurations, with a default value of 6.
Discover and install pre-configured MCP servers in one tap. Powered by [MCPRouter](https://mcprouter.co/), the new Marketplace is a unified directory of integrations—time zones, web automation, image generation, data analysis, and more. Just enter your MCPRouter API key, search by name, and install any server instantly. All endpoints, headers, and auth keys are auto-configured—no manual setup required.
Open and export AI-generated PDFs, images, Office 365 documents, iWork files, videos, and audio.
omlx is now available as a built-in provider in the remote server API key list. omlx is a self-hosted MLX inference server that runs locally on your Mac and exposes an OpenAI-compatible API at localhost:8000/v1. Select omlx from the provider list, and Privacy AI will connect to your local server automatically, letting you chat with any MLX model you host without any cloud dependency.
The built-in on-device model used for internal tasks has been upgraded from Qwen3-0.6B to Qwen3-0.8B. The local Qwen3.5-0.8B model supports text only for the sake of bundle size. You can download the full Qwen3.5-0.8B model from the recommended list with vision support. Larger Qwen3.5 models like 2B and 4B bring improved reasoning and instruction-following quality. You can download either the GGUF or MLX version. Note that MLX versions only support text generation for now.
PaddleOCR-VL-1.5 is now available in the recommended vision model list. This local GGUF model is purpose-built for OCR — it accurately extracts text from images, scanned documents, receipts, and complex visual layouts. Everything runs entirely on-device, so your documents never leave your device.
Privacy AI now automatically detects whether a downloaded MLX model is a language model (LLM) or a vision-language model (VLM), so you no longer need to configure this manually. Simply download any supported MLX model from HuggingFace and the app will identify the correct type and load it accordingly.
GGUF vision models now support downloading mmproj (vision projection) files separately from the main model. After downloading a vision-language model like Qwen3-VL, Ministral-3, or LLaVA, you can configure the vision projection file from the model detail view. Download mmproj files from a custom URL or import them from local storage with full background session support and progress tracking. Files are automatically integrated into the model's manifest and appear in the "Model Files" list. This feature makes it easier to add image understanding capabilities to GGUF models without re-downloading the entire model.
Privacy AI now officially supports Youtu-VL-4B-Instruct, an advanced GGUF vision model built on the DeepSeek2 architecture with Multi-head Latent Attention (MLA) for high-accuracy image analysis. This model requires Metal GPU acceleration and works on real iOS devices with A-series chips (iPhone 16 Pro and newer recommended). Youtu-VL provides superior accuracy for detailed image understanding and analysis tasks, expanding your capabilities for working with visual content.
MLX models now support flexible tool call parsing with external parser injection, enabling them to handle various tool call formats beyond JSON. Qwen format, Ministral format, LiquidAI format, and custom formats. This makes MLX models as capable as their GGUF counterparts when it comes to tool integration.
Privacy AI now officially supports the LFM2.5-1.2B-Instruct-MLX model with proper tool call parsing configuration. This lightweight 1.2B parameter model from LiquidAI runs efficiently on-device with full tool calling capabilities, making it an excellent choice for iPhone users who want fast, capable local AI with tool integration.
Privacy AI now supports Ministral-3-3B-Reasoning-2512, a compact 3.4B parameter reasoning model optimized for edge deployment with a 256K context window. The app includes a flexible tool call parser system using chain-of-responsibility pattern that automatically detects and handles different tool call formats. Ministral uses a unique tool call format which is now fully supported. This enables enhanced multi-step reasoning, mathematics, and scientific problem-solving with full tool integration—all running locally on your device.
sherpa-onnx has been upgraded to 1.12.15, adding support for newer TTS models like MatchTTS and expanding available voice options for higher-quality speech synthesis.
Major CPU performance upgrade for iOS and macOS with ARM64 chunking and Flash Attention optimizations. Adds support for new vision models including Qwen3-VL, CogVLM, and Janus Pro, delivering enhanced multimodal reasoning and lower memory usage on mobile-class devices.
Qwen3-VL, Gemma 3, Qwen 2 VL, Qwen 2.5 VL, and SmolVLM 2 now support tool calling. These vision-language models can act as text-to-text engines, enabling seamless integration with Privacy AI’s local tools and workflows.
Privacy AI now supports the MLX model engine, enabling both text and vision models to run locally. You can directly download models from Hugging Face by entering a repository ID and access token (if required). The new download manager adds resume-on-failure, background downloading, and a model integrity verifier for reliable large-model transfers. MLX models now fully support local tool calls and MCP tool calls, just like GGUF and remote API models. Best of all, MLX model support is included in the Free Plan.
Upgraded to version b6558 with support for Liquid AI series models.
You can now export your usage statistics as a Markdown report directly from the Cost Estimator. Tap the share button to send the report to another app, or copy it to the clipboard. The report is organized into Remote and Local model sections so you can see a clear breakdown of where your usage is coming from.
Privacy AI now supports organizing chats into folders. Create folders up to 3 levels deep to categorize chats by topic, project, or purpose, whether it's separating work discussions from personal queries, grouping research conversations, or organizing learning materials. Navigate through your folder hierarchy with visual breadcrumbs, move chats between folders with an intuitive picker, and create new chats directly inside specific folders.
You can now branch off from any assistant message to explore different conversation paths. A new branch button (arrow icon) appears alongside the copy, edit, and delete actions in assistant messages. Tap it to select a different model, and Privacy AI creates a fresh chat that includes only the conversation history up to that point—perfect for comparing how different models respond to the same context or taking your discussion in a new direction without losing your original thread.
URL citations from web search results now open in Privacy AI's built-in web browser instead of launching Safari. View source content without leaving your conversation, with access to reader mode for distraction-free reading, cookie management for privacy control, and easy URL copying for reference. This keeps your research workflow seamless and focused within a single app.
A new "Rethink" button allows you to resume the AI's thinking process without re-executing tool calls. This is useful when you want the AI to continue reasoning from where it left off without triggering the same tools again. It is perfect for exploring alternative interpretations, or extending analysis without incurring additional API costs from repeated tool executions.
Conversations are now grouped by year and month (e.g., "December 2024", "November 2024") for easier navigation. Each month section includes a delete button that lets you remove all chats from that period at once, making it simple to manage and clean up old conversations.
Code blocks in Markdown now render with syntax highlighting for 40+ programming languages. AI-generated code is easier to read and automatically adapts to light and dark mode for a consistent viewing experience.
Chat history now pulls in only the most relevant excerpts from attached files for your current question, keeping context lean while still recalling what matters.
Control how often text refreshes under throttle to prevent UI lag and battery drain with very fast models.
When cloning or forking a chat with the same model, all settings (temperature, top_p, top_k, context, etc.) are preserved.
Rebuilt history handling now supports binary attachments (images, PDFs, and more) directly in chat history, so you can reference them later.
The Local AI Gateway (OpenAI-compatible local server) now accepts image inputs in chat completion requests. You can pass images as remote URLs. The gateway downloads them automatically, or as base64 data URIs embedded directly in the request. This follows the standard OpenAI `image_url` content part format, so any OpenAI-compatible client can send vision requests to your local server. Works with vision capable models such as PaddleOCR-VL-1.5 (on-device) and GLM-4.6V (remote).
GLM-5 models from Zhipu AI requires a strict tool call whitelist policy and it could not use the auto-tool selector due to missing tool call policy support. With this update, the auto-tool selector correctly meets GLM-5 requirements and it can invoke tools automatically, just like other supported models.
The AI Gateway now helps you monitor and understand your AI interactions better. The Protocol Inspector lets you see detailed logs of all API calls made to AI providers, with easy-to-read formatting and search capabilities to quickly find specific requests. The Token Usage Statistics feature tracks how many tokens each conversation uses, showing you input, output, and total token counts per model in a clean table view. Your usage history is saved across sessions so you can monitor patterns over time. These tools help you debug issues with AI responses and keep track of API costs across OpenAI, Claude, HuggingFace, and other providers.
Privacy AI now fully supports the Open Responses API(https://www.openresponses.org/), providing enhanced conversation chaining capabilities with multi-turn interactions and improved context management.
Privacy AI now officially supports LongCat, Meituan's open-source AI platform featuring advanced models with efficient Mixture-of-Experts (MoE) architecture. Access powerful models like LongCat-Flash-Chat and LongCat-Flash-Thinking for both general conversation and extended reasoning tasks. Each account automatically receives 500,000 Tokens free quota per day
Privacy AI now supports the Min-P (Minimum Probability) sampling parameter for compatible providers including OpenRouter and DeepSeek. Min-P offers an advanced alternative to Top-P (nucleus sampling) by filtering tokens based on a minimum probability threshold relative to the most likely token. This gives you fine-grained control over output quality and diversity—higher values (e.g., 0.1) produce more focused and consistent outputs, while lower values (e.g., 0.01) allow greater creative diversity.
Privacy AI now includes dedicated OpenRouter protocol support with native integration for OpenRouter-specific features. Now you can configure web search capabilities with the :online suffix, choose between native or Exa search engines, and control max results and context size—all through a dedicated settings panel. OpenRouter models are tagged with an "openrt" indicator for easy identification. This extends beyond standard OpenAI compatibility to unlock OpenRouter's full feature set while maintaining the same debugging and inspection tools you're familiar with.
You can now duplicate any remote model configuration with a single tap. The new model will clone all settings: API key, server URL, system prompt, tools configuration, and parameters—are copied from the original. The duplicated model opens immediately for editing, making it easy to create variations.
Privacy AI now supports the HuggingFace Inference API. Use your HuggingFace token to access 5,000+ models with the same debugging experience available for OpenAI and Claude protocols. All outgoing requests are captured by the built-in inspector, and the integration works seamlessly with mitmproxy and Charles Proxy for complete end-to-end API debugging.
Nous Research is now available as a new remote provider. Connect directly to their API to access additional cutting-edge models and integrate them into your workflows just like other cloud providers.
Privacy AI now supports the Claude API protocol, enabling direct connections to Claude models, Z.ai GLM4.6, and Kimi K2 Thinking with full support for multimodal inputs (text, images, PDFs), tool calling, and extended thinking modes. The built-in protocol inspector captures all API requests, responses, and SSE streaming events in real-time, making it easy to debug API interactions and verify behavior. Compatible with Anthropic's official endpoint and any third-party providers supporting the Claude API standard. If you're already a Z.ai or Kimi Code Plan member, you can use your existing API key to connect their endpoints and chat at no extra cost.
Added support for the max_tokens parameter for both local and remote models. You can now precisely control the maximum number of tokens generated in each response through the model settings interface. This parameter is fully integrated with the model configuration system, and works seamlessly with all protocols.
API keys now support multiple named endpoints, allowing you to configure different endpoint URLs under a single API key. This is perfect for services that expose multiple endpoints with the same authentication (e.g., GLM4.6 / Kimi K2 Coding Plan). Each endpoint can have its own name and URL, making it easy to organize and switch between different configurations while reusing the same API credentials. Manage all your endpoints directly in the API Key detail view for streamlined configuration.
Remote model settings now include a Function Call Wait Time parameter that allows you to specify a delay (0-30 seconds) between consecutive tool calls. This helps prevent API rate limiting issues when using models with function calling capabilities. The workflow automatically waits the configured number of seconds between each tool execution, displaying a status message to keep you informed. This is especially useful for providers with strict rate limits or when executing multiple sequential tool calls in complex workflows.
We’ve added Vercel (vercel.com) as a new official provider. Vercel offers some of the most cost-efficient AI models available today. Free-tier users receive $5 in credits every 30 days for use with AI Gateway models — a great way to explore their ecosystem at no cost. All Vercel model pricing data has been fully refreshed within Privacy AI.
Added support for OpenAI’s 'gpt-image-1' and HuggingFace’s 'black-forest-labs/FLUX.1-dev'.
Added OpenAI protocol text-to-speech endpoints, including gpt-4o-tts' and 'gpt-4o-mini-tts'.
When using the Reader's "Chat with Document" feature, a new Select Tools section now appears on the setup screen before you start the session. You can enable any combination of local tools and MCP tools that the AI can use while analyzing your document, giving you full control over what capabilities are available during the conversation.
Privacy AI now supports rendering SVG (Scalable Vector Graphics) images directly in chat messages. SVG files from workspace or URLs are automatically detected and rendered using SVGKit, providing crisp, scalable graphics on both iOS and macOS. This enables AI to display vector diagrams, charts, and illustrations that scale perfectly at any size.
You can now select and import multiple documents at once from the file picker. All selected files are processed sequentially with a smooth progress indicator, making it easier to work with multiple documents without having to import them one by one.
Privacy AI now supports converting Google Books pages into readable text. When you share or paste a Google Books URL, the app automatically detects scanned book page images, extracts text from each page using OCR technology, and formats the content into readable markdown with page numbers. Multiple pages are processed concurrently for faster results, making it easy to extract and analyze content from Google Books.
Privacy AI now reads and processes CSV (Comma-Separated Values) and TSV (Tab-Separated Values) files seamlessly. Share spreadsheet data from Files, Safari, or any app, and the content is automatically parsed and normalized for AI analysis. Multi-line fields and special characters are handled correctly, making it easy to analyze data tables, export reports, or extract insights from structured data files.
CSV and TSV files now display in a dedicated table-based previewer with pinned header rows and smooth scrolling. The table format shows your data in columns and rows just like a spreadsheet, with the header staying fixed at the top while you scroll through up to 100 rows of data.
Now it supports reading and converting contact files (.vcf) and calendar files (.ics) into structured markdown. Share vCard files to view contacts with names, emails, phone numbers, addresses, and other details in an organized format. Import iCalendar files to see events, tasks, and journal entries with dates, locations, attendees, and reminders—perfect for analyzing schedules or exporting calendar data for AI assistance.
The Reader now supports opening ZIP archives and displaying multiple files within. Excel files with multiple sheets are properly handled, allowing you to switch between sheets seamlessly. This is powered by the new ProcessedContentCollection system that handles multiple content items from a single source, making it easy to work with complex documents and compressed archives without manual extraction.
When sharing files to Privacy AI, you can now choose how the content is handled before sending it to the model. Send as is preserves the original file format. Convert to Markdown extracts text and converts it into lightweight Markdown to reduce token usage—recommended for most workflows. Your last selection is remembered across sessions, so you don’t need to reselect it each time.
Privacy AI now supports SRT and VTT subtitle files when sharing content to the app. Both SubRip (.srt) and WebVTT (.vtt) subtitle formats are automatically recognized and processed as plain text, making it easy to analyze video captions, translate subtitles, or extract dialogue from media files. Simply share a subtitle file to Privacy AI from Files, Safari, or any other app, and the content will be ready for AI analysis.
Privacy AI now supports OpenRouter’s native video input protocol. You can send video files directly to supported multimodal models—such as Gemini 2.5 Flash, Flash Lite, Pro, and others—without any manual pre-processing. The app handles encoding and upload automatically, offering a seamless video-to-AI workflow.
Attachments now preview Microsoft Office, Apple iWork, PDF, video, and audio files.
Connect with the Privacy AI community on Discord! We've added a new feedback method that lets you join our Discord server to chat with us, send feedback, get help, and connect with other users. Find the Discord option in Settings -> Feedback, or visit the #general channel directly to start engaging with the community.
The new 'use_clipboard' tool enables AI to read from and write to your system clipboard on both iOS and macOS. Read clipboard content (text, URLs, or detect images), write text or URLs to the clipboard, check content type without reading full content, or clear the clipboard entirely. The tool seamlessly integrates with the existing ClipboardUtil for cross-platform compatibility and supports mode aliases (paste/read, copy/write) for natural language interactions.
Privacy AI's share extension now includes 13 comprehensive quick actions that cover content understanding, writing productivity, and research workflows. New actions include Key Takeaways (extract main points), Explain Simply (ELI5-style explanations), Proofread & Polish (grammar and style improvements), Extract Action Items (identify tasks and deadlines), Create Outline (structured hierarchy), Expand & Elaborate (detailed development), and Research Deep Dive (comprehensive analysis with citations). All 13 actions are organized by frequency of use for faster access.
Long-press the Privacy AI app icon to access Quick Actions. Tap "Quick Ask" to open a streamlined interface for asking questions with real-time streaming responses. Quick Ask uses the same default model configured for Siri in Settings. Perfect for quick queries on the go and will not pollute your chat history.
You can now sync chats and data between local storage and iCloud Drive with dedicated sync buttons in Settings. The sync operation skips existing files to prevent overwrites, provides real-time progress tracking with file counts and current file names, and shows detailed completion statistics. Two-way sync is supported—transfer from local to iCloud or from iCloud to local, making it easy to migrate your data or consolidate chats from different storage locations. And Clean Local and Clean iCloud options are available with double-confirmation warnings to help manage storage space safely.
A new Storage Management section in Settings provides detailed visibility into your app's storage usage. View file counts and total sizes for each storage category. Each category displays its purpose and current storage footprint, with one-tap delete buttons to quickly free up space. Directory information refreshes automatically when switching between local and iCloud storage, and deletion requires confirmation to prevent accidental data loss. This makes it easy to identify large directories and clean up storage without leaving the app.
The Chat and Reader editors now allow pasting images, videos, and files directly from the clipboard. On iPad, you can also drag and drop files or images into the chat editor — a smoother, more natural way to attach media.
Most core features are now permanently free, including Local Models, Apple’s on-device Foundation Model, iCloud sync, natural language chat, Reader, 25+ built-in tools (Search, News, Stocks, Weather, Health, Email, Calendar, etc.), conversation cloning, advanced export, audio/video transcription, Siri & Shortcuts, and more. Only advanced features—Cloud models, MCP, and custom API providers—require a Pro Plan subscription.
Integration with the latest iOS 26 UI and Apple’s on-device Foundation Model.
A new hands-free interface that lets you talk naturally with any model.
The “What’s New” screen now appears automatically after each app update.
Protect conversations and data with biometric authentication. When enabled in Settings → System → Security & Privacy, Privacy AI requires Face ID, Touch ID, Optic ID, or your device passcode to access the app. Authentication is required at launch and when returning from background, with prompts appearing automatically for seamless security.
Questions, feedback, or just want to say hi? We're here.
Available on iPhone, iPad and Mac. No subscription required to get started.
Download Free on App StoreiOS · iPadOS · macOS