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
Ask a question once and get answers from multiple AI models simultaneously. Privacy AI sends your query to a council of models you choose, collecting their independent responses side by side. Use the council to explore different perspectives, stress-test ideas, compare model quality, or simply get richer answers than any single model can provide.
Local Models
Run any GGUF-format language model entirely on your device, with no internet connection required. Privacy AI uses the llama.cpp engine to run models ranging from lightweight 1B-parameter models on older devices to large 70B+ models on high-memory devices. Both text and vision models are supported, so you can chat, analyze images, and use all 50+ tools completely privately.
Local Models
Run MLX-format models optimized for Apple Silicon chips. The MLX engine is purpose-built for the unified memory architecture of M-series chips, making it significantly faster for generation on compatible Mac and iPad hardware. Download any MLX model from HuggingFace and use it with all of Privacy AI's tools and features.
Local Models
Use Apple Intelligence, Apple's built-in on-device model available on iOS 26+ and macOS 26+. No API key or download required; the model runs entirely on-device using Apple's secure on-device framework. Select "Apple Intelligence" as your model to use it in any chat, with all tools and features available.
Local Models
Run Qwen3 TTS (text-to-speech), Qwen3 ASR (speech-to-text), and VAD (voice activity detection) models using the MLX audio engine. All audio processing happens on your device. Use these models for real-time transcription, translation, and natural-voice TTS output with no cloud dependency.
Local Models
Transcribe speech to text in real time using on-device MLX speech recognition and voice activity detection models. Supports 11 languages: English, Chinese, Japanese, Korean, French, German, Spanish, Italian, Portuguese, Russian, and Arabic. Transcription is paired with real-time translation so you can see speech from one language translated to another instantly. Everything runs on your device with no audio ever sent to the cloud.
Local Models
Convert any text to natural-sounding speech using the Kokoro-82M voice model, powered by sherpa-onnx and running entirely on your device. Choose from 53 voice styles across multiple accents and genders. Export generated audio as M4A, WAV, or AIFF files to use in other apps.
A complete professional AI toolkit, all on-device.
Seven ready-made council configurations let you start a multi-model discussion with one tap. Available templates: Debate (structured pros/cons for decision making), Research (deep analysis across multiple perspectives), Balanced (mixed models for general-purpose discussions), Business Strategy (strategic planning and market analysis), Critical Thinker (challenge assumptions and stress-test ideas), Game Theory: Classic Games (explore strategic scenarios from classic game theory), and Game Theory: Custom Scenario (describe your own strategic situation for the council to play out).
Two discussion modes control how council members respond to your query. Standard mode sends the same query to all members simultaneously and they each respond independently. Role Assignment mode has the chairman first analyze your query, then assign each member a unique perspective or analytical role so the group produces maximally diverse and complementary insights.
Council members play through strategic negotiation scenarios across multiple rounds, inspired by the research paper "Can Large Language Models be Good Game Theorists?" (arXiv:2411.05990). In Classic Games mode, models compete or cooperate in well-known game theory scenarios. In Custom Scenario mode, you describe any strategic situation and the council acts it out, revealing how different decision-making strategies unfold over time.
Privacy AI maintains a personal memory store that the AI can recall during conversations. Add, edit, and delete individual memories at any time from the Memory settings screen. Memories are stored privately on your device and synced via iCloud.
Memories are organized into sectors such as life, work, and study, making it easy to browse and manage your stored information by category. Create memories in the relevant sector so the AI retrieves the most context-appropriate information.
When memory is enabled for a chat, Privacy AI automatically searches your memory store and injects the most relevant memories as context before every AI response. The AI uses your personal background, preferences, and prior knowledge to give more relevant, personalized answers.
After a conversation, tap the import button to extract key facts and preferences from the chat and save them to your memory store with one tap. Privacy AI uses AI to identify what is worth remembering and formats it automatically.
Search your stored memories using natural language rather than exact keywords. Type what you are looking for and Privacy AI finds the most semantically similar memories, even if they use different words.
Analyze images and photos using local vision models such as Qwen-VL and PaddleOCR-VL. Drop an image into your chat and the model will describe it, extract text from it, answer questions about it, or use it as context for further reasoning. All image understanding happens on your device, with no images ever sent to a server.
Run vision-capable MLX models for image understanding on Apple Silicon devices. These models leverage the fast MLX engine for efficient image plus text reasoning, letting you attach photos to your messages and get intelligent analysis entirely offline.
Transcribe spoken audio to text using Whisper models powered by WhisperKit. Download a Whisper model once and use it to transcribe voice notes, recorded meetings, or any audio file, completely offline. Multiple model sizes are available to balance speed and accuracy.
Local models can call any of the 50+ built-in tools, from web search and calendar management to image generation and shell commands. When you ask a local model to look something up or perform an action, it will use the appropriate tool automatically, then return the result as part of its response, all without leaving your device.
Create a reusable template for a local model with a customized parameter set and tool selection. Every chat created from that template inherits the same settings without needing to reconfigure them. Configurable parameters include: context window size, seed, temperature, mirostat mode, Direct I/O, BOS/EOS tokens, batch size, top_k, top_p, min_p, repeat penalty, repeat last N, max tokens, tool result limit, image support, and more.
Browse and download GGUF and MLX models directly from HuggingFace inside the app. Downloads resume automatically if interrupted and show real-time progress. Large models with separate vision projection files (mmproj) download both files together and link them automatically.
Download a model once and it becomes available on all your devices via iCloud Drive. Switch between iPhone, iPad, and Mac without re-downloading large model files. Model settings and configurations also sync across devices.
Import a GGUF model file you already have by picking it from the Files app. Privacy AI copies it to its model storage and makes it immediately available for chat. Useful for models obtained outside HuggingFace or for offline device setups.
For vision models that require a separate projection file (mmproj), Privacy AI handles downloading and linking the file automatically. You can also import a local mmproj file manually if you have it on your device.
Upload any audio or video file from the Files app or photo library and transcribe it to text on your device. Export the transcription as an SRT subtitle file, a VTT file, or plain text. Useful for transcribing meetings, lectures, podcasts, or voice memos.
Run Qwen3-based TTS, ASR, and VAD models using the MLX audio engine for Apple Silicon devices. These models offer high-quality speech synthesis and recognition while leveraging Apple Silicon's performance for fast, low-latency processing.
Tap the speaker icon on any AI reply to have it read aloud with your chosen TTS engine. Supports Apple's built-in voice, the local Kokoro engine, and MLX audio models. Adjust speed and voice in settings.
Tap the microphone icon in the chat input field to dictate your message using the device's built-in speech recognition. Dictated text appears directly in the input field for editing before sending.
Use OpenAI-compatible TTS endpoints (such as gpt-4o-tts) as a cloud voice provider alongside the local TTS engines. Configure a cloud TTS provider in settings to use it for Read Aloud on any chat.
Trigger Privacy AI through Siri using natural language. Say "Ask Privacy AI to..." to open a new chat with your query. Configure which tools are enabled for Siri-triggered sessions in settings.
Connect to any of the supported cloud AI providers using your own API key. Supported providers: **LLM providers:** OpenAI (GPT-4o, o3, o1), Anthropic (Claude 3.5/3.7/4), DeepSeek, xAI (Grok), OpenRouter (300+ models), Gemini (Google), Perplexity, Mistral, MiniMax, HuggingFace, Kimi (Moonshot), Groq, Z.ai (GLM), Nvidia, GitHub, Vercel, Nous Research, LongCat. **Self-hosted servers:** Ollama, LM Studio, llama.cpp server, vLLM, omlx, LocalAI, Jan AI, Privacy AI Gateway. **Multimedia:** ElevenLabs (voice), Replicate (image/video/voice), Fal.ai (image/video), Stability AI (image/music), Runway ML (video/image).
Create a reusable connection template for any remote model with preset tools and settings. All chats started from that template share the same model, system prompt, tool configuration, and parameter settings. Clone or duplicate templates to quickly create variations without starting from scratch.
When creating a remote model, Privacy AI shows the latest input and output token prices so you can evaluate cost before committing. Compare the same model across providers to find the cheapest option. Context window size is automatically filled in from model information to save setup time.
Privacy AI speaks four AI protocols natively: OpenAI Chat Completions (used by OpenAI and most compatible providers), OpenAI Responses API (conversation chaining with persistent threads), Anthropic Claude API (native Claude protocol with extended thinking support), and HuggingFace Inference (direct HuggingFace model access). Each protocol unlocks its provider's specific capabilities such as tool use, streaming, vision, and reasoning.
Each API provider can be configured with multiple named endpoint URLs, for example a standard endpoint and a separate coding-optimized endpoint. Switch between endpoints without creating duplicate API key entries. Useful for providers like Z.ai that offer different endpoints for general and coding tasks.
Connect to any OpenAI-compatible self-hosted server running on your local network or a custom URL. Supported servers include Ollama, LM Studio, vLLM, LocalAI, Jan AI, llama.cpp server, and omlx. Just enter the server address and port; no API key is required for most local servers.
Privacy AI can run a local OpenAI-compatible HTTP server on your device. Any OpenAI API client can connect to this server and use either local models (running on-device) or remote models (forwarded through Privacy AI). Build custom integrations, connect other apps, or use Privacy AI as a proxy for your personal model infrastructure.
The AI Gateway accepts image inputs, allowing any client that connects to it to send images for visual analysis by local or remote vision models.
Privacy AI tracks token usage per model and provider over time. View cumulative input and output token counts, understand which models you use most, and monitor usage trends.
Privacy AI runs natively on all iPhone and iPad models. On iPhone, the interface is optimized for one-handed use with a compact layout. On iPad, the wider sidebar shows more content, layouts adapt for landscape orientation, and up to 12 parallel conversations can be open at once. All features are available on both form factors.
Privacy AI runs as a native macOS app with full macOS window chrome, menu bar integration, keyboard shortcuts, and resizable windows. The interface follows macOS design conventions rather than forcing iOS navigation patterns. All local model inference, tools, and features work on macOS just as they do on iOS.
Choose Light, Dark, or Auto theme in settings; Auto follows the system appearance. Privacy AI also supports the dark mode app icon on iOS 18+ to match your Home Screen aesthetic.
Configure detailed inference parameters per model. For local models: context window, seed, temperature, mirostat mode, Direct I/O, BOS/EOS tokens, batch size, top_k, top_p, min_p, repeat penalty, repeat last N, max tokens, tool result limit, and image support. For remote models: context, seed, temperature, top_p, min_p, presence penalty, frequency penalty, max tokens, reasoning effort, streaming toggle, parallel tool calls, max tool calls per turn, interleaved thinking, tool result limit, and usage tracking.
Enable Auto Tool Selector to let the AI automatically choose the best tools for each query from all available local and MCP tools. Instead of manually configuring a tool set per model, the AI reads the available tools and picks the most relevant ones at query time.
Adjust the visual appearance of chat messages: choose font size, font color, and background color. Select message display mode (Typewriter for streaming character-by-character, Chunked for paragraph-at-a-time, or Immediate for instant full-response display). Set the refresh rate from 5 to 60 FPS for streaming responses.
Configure a primary search engine and one or more fallback engines for the web search tool. Privacy AI tries the primary engine first and falls back automatically if it fails. Choose from the supported search providers in settings.
View a breakdown of storage usage by category: AI models, chat history, and workspace files. See exactly how much space each category occupies and tap to clean up unused or old files with a single action.
Set a default system prompt for any model or for an individual chat. The system prompt defines the AI's behavior, persona, or constraints for that context. Override the model-level default with a chat-specific prompt when needed.
Share any supported document, image, URL, or text from any app to Privacy AI using the iOS Share Sheet. Privacy AI's Action Extension receives the shared content and opens a new chat session with it as context, letting you immediately ask the AI to summarize, translate, explain, or otherwise process what you shared.
When sharing content to Privacy AI, choose from your saved prompt templates, type a custom prompt, or select from built-in quick actions such as Summarize, Translate, and Explain. This lets you direct the AI's first response without needing to type in the chat.
Choose which model (local or remote) processes the shared content, and select which chat folder the resulting conversation is saved to. All configured models and folders are available in the Action Extension without opening the main app.
Choose whether to send the shared file to the model as-is or to convert it to Markdown first using Privacy AI's built-in document converter. For translation actions, select the target language before sharing.
Enable or disable specific tools for the Action Extension chat session. Keep the tool set focused for quick actions, or enable all tools for complex processing tasks.
Chats, memories, prompt templates, model configurations, and settings sync automatically across all your devices via iCloud. Start a conversation on iPhone and continue it on iPad or Mac without any manual transfer.
Paste text, images, or files from the clipboard directly into the chat input field. Privacy AI detects the clipboard content type and handles it appropriately, inserting text as a message draft or attaching images and files.
Extract text from images using on-device vision. Paste or attach an image containing text and Privacy AI can read and use the text as context, useful for photos of signs, documents, screenshots, or whiteboards.
Enable Natural Talk mode for a hands-free voice conversation with any model. Privacy AI listens for your voice, transcribes it on-device, sends it to the AI, and reads the reply aloud automatically. No manual tap required to start or stop listening.
Long-press the Privacy AI app icon on the Home Screen to access Quick Ask, a streamlined interface that opens directly to a new chat input with no navigation steps. Ideal for rapid questions when you need an answer fast.
After each app update, Privacy AI shows a summary of what changed in the new version. The screen appears automatically on first launch after an update and can be dismissed with a single tap.
View usage statistics broken down by model and provider with charts. See token consumption over time, identify which models you use most, and understand your AI spending patterns at a glance.
Export your usage report as a Markdown document to share with a team, include in expense reports, or archive for your records.
The Free plan includes all local model features (GGUF, MLX, Apple Intelligence), iCloud sync, over 50 built-in tools, all chat features, export, offline transcription, Siri integration, and the Reader. No subscription required to use Privacy AI with local models and on-device AI.
The Pro plan adds access to cloud AI providers (OpenAI, Anthropic, DeepSeek, and all others), MCP server integration, and the ability to configure custom API providers with your own keys. Required for any cloud-based model or MCP functionality.
Pin frequently used chats to the top of your chat list so they are always one tap away. Pinned chats stay at the top regardless of how many other conversations you create.
Chats are automatically organized into year-month groups in the sidebar, making it easy to find a conversation from a specific time period without manual organization.
Create folders to organize your chats by project, topic, or any system you prefer. Folders can be nested up to three levels deep. Move chats between folders by dragging or using the move menu.
Lock Privacy AI with a passcode so your conversations and memories are not accessible without authentication. Useful on shared devices or for extra privacy when the app contains sensitive information.
Long user messages are automatically collapsed to save screen space, showing just the first line with a tap-to-expand option. Previous AI responses are also collapsed so only the current response is fully visible. This keeps the chat focused without losing access to earlier content.
Every tool the AI used during a conversation is recorded in a collapsible history panel. Tap any entry to see which tool was called, what parameters were sent, and what data was returned. Useful for understanding exactly what the AI did to produce its answer.
Tap the speaker icon on any AI reply to have it read aloud. Privacy AI supports multiple TTS engines: Apple's built-in speech synthesis, the offline Kokoro-82M engine via sherpa-onnx, and MLX audio models. Choose your preferred voice in settings.
Ask the AI to reconsider its last answer without rerunning all the tool calls it already made. The AI regenerates its response using the same context and tool results but may reason differently, useful when the first answer missed the point.
Resend the current query to get a fresh response. The AI runs the full query again, including any tool calls, giving you a new take on the same question.
Attach files and media to any message in multiple ways: paste images directly from the clipboard, pick up to 10 images at once from the photo album, select any file type from the Files app, draw an image with Apple Pencil or your finger, take a photo with the camera, or pick a video from the album. Attachments can be sent to the model as-is, or Privacy AI's built-in converter can transform nearly any file type (video, audio, photos, PDFs, Word, Excel, PowerPoint, EPUB, and more) into Markdown for token-efficient processing. Toggle conversion on or off per attachment before sending.
Tap any attachment thumbnail to preview the full file before sending. Supported preview types include text, Markdown, CSV, PPTX, DOCX, PDF, PNG, JPG, SVG, EPUB, and more.
Copy the full content of any chat to the clipboard with one tap. The copy respects current display options so you get exactly what you see.
Duplicate any existing conversation to start a new branch from it. Useful for exploring different follow-up directions from the same starting point without losing the original thread.
Open multiple chats simultaneously and switch between them without losing streaming progress. Privacy AI supports up to 8 parallel conversations on iPhone and 12 on iPad, so you can run long-running queries in one chat while working in another.
Tap any previously sent message to edit it and resubmit. The conversation is rewound to that point and the AI responds to the edited version, letting you refine your question without starting over.
A visual indicator shows how much of the model's context window is currently in use. When context fills up, older messages are automatically summarized or dropped according to the model's configuration.
Your in-progress message is automatically saved if you switch away from the chat or close the app. When you return, the draft is restored exactly as you left it.
Drag files and images from any app directly into the Privacy AI chat input field. iPad's multitasking support makes it easy to pull a document from Files or an image from Photos into your conversation without switching apps.
Select up to 10 images at once from the photo picker to attach to a single message. All selected images are previewed before sending and can be individually removed.
Export a chat to plain text, Markdown, HTML, JSON, PDF, EPUB, or as an image. Before sharing, choose whether to include user queries, AI thinking content, tool call history, and attachments. Preview the rendered result before saving or sharing.
When a model produces thinking or reasoning content (such as DeepSeek-R1, OpenAI o1/o3, or Gemini thinking), Privacy AI shows it in a collapsible section above the final answer. Tap to expand and read the full step-by-step reasoning process.
Privacy AI natively supports thinking and reasoning models that show their step-by-step thought process before answering. Compatible with DeepSeek-R1, OpenAI o1/o3, Gemini thinking models, and local models with thinking support. Interleaved thinking is preserved across tool calls so the full reasoning chain is always visible.
AI replies are rendered with a full-featured Markdown engine. Supports headings, lists, tables, bold, italic, code blocks, LaTeX math expressions, and syntax highlighting. Font size, font color, and background color are all adjustable in settings.
Reference links in AI responses are rendered as rich preview cards rather than plain URLs. Tap a citation to open a mini browser preview inside the chat, showing a summary of the linked page without leaving Privacy AI.
HTML, CSS, and JavaScript code blocks can be rendered live in a built-in WebView. Tap the preview button on any HTML code block to see it rendered, and export a full-length screenshot of the rendered result.
SVG markup in AI responses is rendered as a live vector graphic inline in the chat. Useful for diagrams, charts, and illustrations generated by the AI.
Mermaid diagram code blocks (flowcharts, sequence diagrams, Gantt charts, and more) are automatically rendered as visual diagrams. No external tools needed.
Code blocks are syntax-highlighted for all common programming languages, making AI-generated code easier to read at a glance.
Write, save, and name prompt templates for tasks you repeat often. Open a template from the prompt picker in any chat to instantly paste the full prompt text into the input field. Edit or delete templates at any time.
Privacy AI includes a curated library of community-created prompts from the awesome-chatgpt-prompts project. Browse by category to find useful starting points for coding, writing, roleplay, analysis, and more.
Pin your most-used prompt templates to the top of the prompt picker so they are always the first options you see, regardless of how many templates you have saved.
Paste any URL to import the full text of a web page as clean Markdown, stripped of ads and navigation. Paste multiple URLs at once for batch processing. For pages that require JavaScript to load content, enable JS rendering for accurate extraction.
Import documents from the Files app and Privacy AI converts them to readable text automatically. Supports PDF (including scanned PDFs with OCR), Word (.docx), Excel (.xlsx), PowerPoint (.pptx), CSV, EPUB, and many more formats. The converted text is used as context for your chat.
Extract text from scanned or image-based PDFs using on-device optical character recognition. No internet connection is needed, and the extracted text is immediately available as chat context.
Attach multiple files of different types in a single message. Combine a PDF, a spreadsheet, and an image in one query for unified cross-document analysis.
Paste a list of URLs and Privacy AI fetches and converts all of them in one operation. Useful for researching a topic across several sources or comparing content from multiple pages.
Long-press any document in the Files panel to launch a document-focused chat with a pre-configured tool set. The document is loaded into context and the AI is ready to answer questions, summarize, translate, or extract information from it.
Every chat has its own persistent file workspace. The AI can save, read, edit, and list files in this workspace across the entire conversation. Workspace files survive app restarts and are accessible by all file tools. Use the workspace to store intermediate results, generated code, or any file the AI produces during a long task.
Files too large to fit in context are automatically split into chunks. Privacy AI summarizes, translates, or analyzes each chunk and aggregates the results, so large documents can be processed even on devices with limited context windows.
Open ZIP files containing multiple documents. Privacy AI extracts the contents and makes each file available for individual processing or batch analysis.
Import contact files (.vcf) and calendar files (.ics) to let the AI read and reason about contact information or calendar events from any source.
Read Excel workbooks that contain multiple sheets. Privacy AI presents each sheet's content for analysis, letting the AI work across an entire workbook rather than just the first tab.
Open and read .gz compressed files directly. Privacy AI decompresses the file on the fly and makes its contents available as chat context, no manual decompression needed.
Paste a YouTube URL to extract the video's captions or auto-generated subtitles as text. Use the transcript as context for summarization, translation, Q&A, or any other analysis.
Describe an image in plain language and Privacy AI generates it using your configured image provider (Replicate FLUX, Fal.ai, Stability AI, Runway ML, or OpenAI). Generated images appear inline in the chat and can be saved, shared, or used as input for follow-up edits.
Draw and edit images directly inside Privacy AI using Apple Pencil or your finger. Annotate photos, sketch diagrams, or create artwork to attach to a chat or refine an AI-generated image.
Rotate, flip, and zoom AI-generated images directly in the chat. Interact with the image before saving or sharing it.
Generate short video clips from text prompts or image inputs using Runway ML, Fal.ai, Replicate, or MiniMax. Videos play inline in the chat and can be exported to the photo library or shared directly.
Generate realistic speech audio from any text using ElevenLabs or Replicate voice models. Choose a voice style, generate the clip, and save it as an audio file or share it from the chat.
Describe the music you want, including genre, mood, tempo, and instrumentation, and Privacy AI generates a matching audio clip using Replicate MusicGen, MiniMax, or Stability Audio.
Tools for finding information on the web, news sources, academic databases, and specialized catalogs. Use these to answer questions about current events, research, places, books, games, and more.
Search the web and return ranked results. Choose Speed mode for fast lookups, Balance mode for mixed quality and speed, or Quality mode for the most thorough search. The AI uses web results to answer questions that require current or real-world information.
Fetch recent news articles from multiple news providers including NewsData.io, NewsAPI, GNews, and Currents. The AI retrieves headlines and summaries for any topic, company, or keyword you ask about.
Retrieve Wikipedia articles, summaries, and related topics. The AI uses Wikipedia to answer factual questions, explain concepts, or provide background context on people, places, and events.
Search academic papers on arXiv by keyword, author, or topic. The AI returns paper titles, abstracts, and links, giving you a research-grade starting point for technical questions.
Browse and read current HackerNews stories and comment threads. The AI can surface top stories, fetch full discussion threads, and summarize community reactions to tech news.
Search for podcasts and episodes by topic or keyword, and retrieve transcripts when available. Useful for finding expert commentary, interviews, or discussions on specific subjects.
Find nearby or distant points of interest including hotels, restaurants, gas stations, and other venues using location-based search. The AI returns names, ratings, addresses, and contact information.
Search Google Books for titles, authors, and subjects. The AI retrieves book metadata, table of contents, bestseller lists, and descriptions to help you find or learn about books.
Search the Steam game catalog for titles, details, reviews, and news. The AI can look up release dates, descriptions, system requirements, and recent updates for any Steam game.
Tools for retrieving financial market data, cryptocurrency prices, exchange rates, and prediction market probabilities. Use these to research investments, track markets, or get live pricing during a conversation.
Retrieve real-time stock quotes, company profiles, financial statements, and technical analysis indicators. The AI can fetch data for any publicly traded company and analyze it in the context of your question.
Look up cryptocurrency prices, market cap, volume, and historical price data for any coin or token. The AI can compare assets, explain price movements, and answer questions about crypto markets.
Fetch current foreign exchange rates between any pair of currencies. Useful for travel planning, international pricing, or currency conversion in calculations.
Search prediction market listings and current probabilities on Polymarket. The AI can retrieve market odds, volume, and resolution criteria for any forecasting question.
Tools for analyzing tabular data, running statistical and mathematical computations, and generating charts. Attach a spreadsheet or describe your data and the AI can summarize it, find patterns, run regressions, or visualize results inline.
Analyze tabular or CSV data with descriptive statistics, value counts, outlier detection, and correlation analysis. Attach a data file and ask the AI to summarize it, find patterns, or answer specific questions about the data.
Compute correlation matrices and run principal component analysis on tabular datasets. The AI returns structured results and can explain which variables are most strongly related.
Run linear regression analysis on tabular data and return model coefficients, R-squared, and residuals. Useful for finding linear relationships and building simple predictive models.
Evaluate mathematical expressions with support for variables, functions, and units. The AI can solve arithmetic, algebra, and scientific calculations precisely.
Generate bar, line, scatter, and multi-series charts from data you provide. Charts are rendered inline in the chat and can be saved or shared as images.
Break a complex problem into multiple structured reasoning steps, showing the AI's thinking process as a visible step-by-step plan before arriving at a conclusion.
Tools that interact with device features and Apple platform services including location, health, calendar, reminders, contacts, and messaging. These tools let the AI take real-world actions on your behalf, from scheduling a meeting to sending a message.
Retrieve the current date and time, and perform date arithmetic such as calculating the number of days between two dates or finding a date N days from now.
Get the device's current GPS coordinates and resolve them to a readable address. Used by weather, directions, and places tools automatically.
Fetch current weather conditions and a 16-day forecast for any location. Returns temperature, precipitation, wind, humidity, and other conditions.
Read Apple Health data including step count, heart rate, sleep duration, nutrition, workout history, and more. All data stays on your device and is never transmitted anywhere.
Get turn-by-turn driving, walking, or transit directions between two locations. The AI returns route steps, distance, and estimated travel time.
Create, query, modify, and delete calendar events in Apple Calendar. The AI can schedule meetings, check your availability, move events, or list upcoming appointments.
Create, query, modify, and complete reminders in Apple Reminders. The AI can set deadlines, mark items done, and list reminders by list or date.
Create, list, and remove alarms on your device. Ask the AI to set a wake-up alarm, a cooking timer, or any other time-based alert.
Compose and send emails through the device's configured mail accounts. The AI drafts the message from your description and sends it on your behalf.
Send SMS text messages to contacts. The AI composes the message based on your instructions and sends it using the device's messaging capability.
Search your device contacts by name, phone number, or email address. The AI retrieves contact details to use in other actions or to answer questions about your address book.
Execute any iOS Shortcut from the Shortcuts app by name. Connect Privacy AI to any automation you have already built, letting the AI trigger complex device actions.
Tools for reading, writing, editing, and searching files in the chat workspace and device filesystem. Use these to give the AI persistent file storage across a conversation, run shell commands, and fetch content from URLs.
Read, write, append to, list, and delete files in the chat's persistent workspace across the entire conversation. Files survive app restarts. The AI uses this tool to save progress and share data across multiple steps of a task.
Read and write files in allowed directories outside the workspace. Configure which directories are accessible in settings for controlled file system access.
Read the contents of a workspace file, optionally specifying a line range for large files.
Make precise string replacements in workspace files without rewriting the whole file. The AI uses this for targeted code edits or document revisions.
Execute shell commands directly. On iOS, over 150 commands are available in a sandboxed environment. On macOS, a full interactive zsh session is available. Useful for file manipulation, data processing, and running scripts.
Search workspace files for patterns using regular expressions, across files up to 10MB. The AI uses this to locate specific content in large documents or codebases.
Fetch the content of any URL or workspace file and return it as text. The AI uses this to read web pages, APIs, or local files as part of a task.
Tools for running code, querying databases, and managing developer infrastructure. Use these to execute scripts, inspect database schemas, or pull App Store Connect data directly in a conversation.
Execute JavaScript code (ES6+) and return the result. Supports file-based scripts and complex computations. The AI can use this to process data, transform formats, or run algorithms.
Connect to MySQL or PostgreSQL databases, run queries, and inspect schemas. Configure database credentials in settings; the AI can then query your databases on demand.
Access App Store Connect data including app listings, user reviews, TestFlight builds, in-app purchases, and sales reports. Sales reports support GZip and TSV formats from the API.
Tools for searching and managing the personal memory store. The AI uses these automatically when memory is enabled, or you can invoke them explicitly to save facts mid-conversation or clean up outdated entries.
Perform a semantic search across your stored memories for any topic or keyword. The AI uses this automatically when memory is enabled, or you can trigger it explicitly.
Add, delete, and list memories with hashtags for organization. The AI can save new information to your memory store mid-conversation for future recall.
Review and clean up your memory store by identifying duplicates, outdated entries, and poorly organized items. The AI suggests and applies tidying actions with your approval.
Tools for generating images, video clips, voice audio, and music using cloud AI services. Ask the AI to create any media from a description, with results appearing inline in the chat ready to save or share.
Generate images from a text description using cloud image generation services including Replicate FLUX, Fal.ai, Stability AI, Runway ML, and OpenAI. Generated images appear inline in the chat.
Convert text to speech audio using ElevenLabs or Replicate. Choose from a wide range of voices and accents, and save the output as an audio file.
Generate short video clips from text prompts or images using Replicate, Fal.ai, Runway ML, or MiniMax. Videos play inline in the chat and can be saved to the photo library.
Generate music tracks from a text description using Replicate MusicGen, MiniMax, or Stability Audio. Describe the genre, mood, and instrumentation and the AI creates a matching clip.
General-purpose tools for clipboard management, note creation, social media, and tool discovery. Use these to move content between Privacy AI and other apps or to let the AI post on your behalf.
Paste text from the clipboard into the AI context, copy AI-generated content to the clipboard, check clipboard contents, or clear the clipboard.
Create a new note in Apple Notes with content the AI generates. Useful for saving summaries, lists, or any AI output directly to your notes app.
Compose and post a tweet to Twitter/X. The AI drafts the tweet from your description and posts it using your configured Twitter account.
List all available tools and get a description of what each one does. The AI uses this to discover what capabilities are available when needed.
Tools for launching sub-agents and managing multi-step task plans. Use these when a task is too complex for a single AI turn, such as deep research that requires many steps, or a project that needs to be tracked across a long conversation.
Launch an isolated sub-agent to tackle a complex sub-task independently and return a complete result. Used for deep research, long-form writing, or any task that benefits from a clean, focused AI session.
Create and track multi-step plans for complex tasks. The AI records each planned step and marks them complete as the task progresses, giving you visibility into what has been done and what remains.
Connect Privacy AI to external tool servers using the Model Context Protocol (MCP). Any MCP-compatible server immediately adds its tools to the AI's available tool set. Use MCP to connect Privacy AI to your own databases, home automation systems, code repositories, or any custom service you build.
Browse and install MCP servers from the MCPRouter marketplace directly inside Privacy AI. Discover community-built integrations for popular services and add them with a few taps, no manual server configuration required.
MCP server connections support Authorization headers for secure access to remote or private MCP servers. Configure bearer tokens or custom authentication headers per server in settings.
Run a fully interactive terminal session on iOS with over 150 built-in shell commands including standard Unix utilities, text processing tools (grep, awk, sed, cut, wc), compression tools (gzip, gunzip), network tools (curl), and more. The terminal is sandboxed to the chat workspace for security.
On macOS, Privacy AI provides a full interactive zsh shell session with unrestricted access to the system. Run any command, execute scripts, and interact with the full macOS environment from within the app.
Inspect the exact request and response for every AI API call made during a conversation. Supports all protocols: OpenAI Chat Completions, OpenAI Responses API, Anthropic Claude API, HuggingFace Inference, and OpenRouter. View the raw JSON payload sent to the server and the raw response received, then copy or share the data for analysis, debugging, or bug reports.
Route all AI API traffic through an external proxy by configuring a host and port in settings. Compatible with mitmproxy, Charles Proxy, and any other HTTP/HTTPS proxy tool. Useful for debugging API calls, inspecting traffic, or working in corporate network environments that require a proxy.
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
Privacy AI