Free 7-day trial · No account required · Perpetual license
The architecture
Other apps wrap a chat box around an API. QARK gives you an agent runtime — composable, controllable, yours.
Build agents that think. Chain agents that execute. Watch them work together in real-time.
Cmd+Alt+Space. A floating overlay from any app. Highlight, transform, paste back — without leaving your workflow.
Advanced retrieval, intelligent chunking, multi-stage reranking, cited sources. Not just search — comprehension.
Pre-configured LLM pipelines: system prompt → model → post-processing → output action. One click, every time.
Branch, fork, search, export. Groups, split view, batch ops. Your conversations are a knowledge base.
Bring your own keys
From frontier models to local inference — switch mid-conversation, compare outputs, control the spend.
Anthropic Claude 
OpenAI GPT / o-series 
Google Gemini 
Groq Ultra-fast 
DeepSeek Reasoning 
Ollama Local models 
LM Studio Local models 
OpenRouter Any model 
Together AI Open source 
xAI Grok 
Perplexity Search 
Cohere Embeddings 
Voyage AI Embeddings 
Jina AI Reranking Everything you need
Every feature. Day one. No tiers, no waitlist.
Explore all features in detail →
Composable agents that call other agents as tools — nested execution with full tracking. Each gets its own system prompt, model, temperature, and context strategy. Build a Research Agent, Writing Agent, and a PM Agent that dispatches both.
Cmd+Alt+Space. A floating overlay from any app. Pick a Flow, transform text, paste back.
Anthropic, OpenAI, Google, Groq, DeepSeek, Ollama, OpenRouter, Together, xAI, Perplexity, and more — every frontier model in one app. Switch providers mid-conversation without losing context.
Vector search, hybrid retrieval, and reranking. Every answer cites its source.
Pre-configured LLM pipelines: system prompt → model → post-processing → output. "Summarize in 3 bullets" → clipboard. "Review code" → diff view. Chain with Sparks for instant text transforms.
Smart groups with custom icons and colors, auto-grouping, auto-naming via local LLM. Fork from any message, branch navigation, split view with resizable panes. Full-text search, batch ops, archive, export to Markdown/HTML/PDF.
Search the web from any conversation — via Tavily, Brave, Exa, Jina, Perplexity Sonar, Valyu, Parallel, or your own local browser for completely free, zero-API searches. Built-in provider search via Anthropic, OpenAI, and Gemini grounding too. Results are parsed, cited, and fed into context.
Generate images with DALL·E, Gemini Imagen, and xAI Aurora — and videos with Sora, Gemini, and xAI — inline in any conversation. Preview, iterate on prompts, upscale, download or share directly. Multiple providers for style, quality, and cost.
Voice in, voice out. macOS native or OpenAI Whisper.
Execute shell commands from within conversations — safe, sandboxed, with full output captured in context. Run scripts, query databases, process files, and pipe results directly into your AI workflow.
Fetch and parse any URL — articles, docs, API responses. Uses Jina Reader, Tavily Extract, Parallel, Valyu, Ollama, or your own local browser to extract content for free with no API key. Cleaned and added to context automatically.
Attach PDFs, DOCX, images, spreadsheets, and code files — everything is parsed, chunked, and indexed for instant retrieval. Drag and drop or paste from clipboard.
Frontier to open-source. Switch models mid-conversation.
Every conversation, every setting, every attachment — stored locally in a SQLite database on your Mac. No cloud sync, no telemetry, no tracking. API calls go direct from your machine to the provider. QARK never sees your data.
Connect directly to each provider with your own API keys. Pay only for what you use — no markup, no middleman, no subscription. Full cost transparency across every conversation.
Run Ollama and LM Studio models completely offline on your hardware. Full privacy, zero cost, no API key needed. Perfect for air-gapped environments and sensitive work.
Set spending limits per provider, per day, or per month. Real-time cost tracking per message shows exactly what each response costs. Automatic alerts when you approach limits. Full cost breakdown across all providers in one dashboard.
All API keys encrypted on disk with a randomly generated encryption key unique to your installation. Keys are never stored in plaintext and never sent to our servers.
Connect external tools, databases, and services via the Model Context Protocol. Extend QARK's capabilities with any MCP-compatible server — from GitHub to Slack to your own custom tools. The agent runtime becomes infinitely extensible.
A local language model runs on-device for auto-naming conversations, smart tagging, and intelligent organization — zero API calls, zero latency, zero cost.
Command palette, split view, tabs, multi-pane layouts, and deep keyboard navigation. Every action has a shortcut — new conversation, switch model, toggle sidebar, jump to search. Vim-style motions for power users. Designed for people who never touch the mouse.
Six ways to manage context — truncation, summarization, or hybrid. Per-conversation control.
Teach QARK reusable abilities — modular capabilities that snap in, stack up, and level up your agents. Custom tools that persist across sessions and share across workflows.
Conversations end. Context shouldn't. Persistent memory gives your agents recall across sessions, projects, and weeks — they remember what matters so you don't have to repeat yourself.
QARK in your pocket. Same agents, same privacy, same workflows — synced through iCloud. Your cloud, not ours.
No subscriptions. No credit packs. Use your own API keys — save hundreds per year compared to ChatGPT Plus or Claude Pro.
Everything you need to replace every AI app on your desktop. Full power, no compromises.
Everything included:
Lock in lifetime updates forever. No renewals, no surprises.
Everything in Standard, plus:
Year 2+ renewal: $19/year for updates only (optional — app keeps working forever)
Everything you need to know about Qark.
Qark is an AI agent kernel — not a chat wrapper. Every interaction dispatches an agent with its own model, tools, system prompt, and context strategy. Agents can call other agents as tools, creating composable pipelines and nested workflows. Built with Rust and Tauri for native performance.
Subscription chat apps lock you into one provider and charge $20/month whether you use it or not. Qark connects to 13+ providers with your own API keys — you pay only for tokens you actually use. Plus you get agents, RAG, Flows, Sparks, MCP, and tools that no chat app offers.
Not at all. Start a conversation, pick a model, and chat — it works like any AI app. The advanced features (agents, flows, RAG, MCP) are there when you want them, but never in the way. Lawyers, writers, scientists, and students use Qark alongside developers.
macOS today, with Windows and Linux support launching within 30 days. An iOS companion app is also in development for on-the-go access with the same agents and privacy.
Most BYOK apps are chat wrappers with a model picker. Qark is an agent runtime — every interaction dispatches a configurable agent that can call other agents as tools, creating nested workflows. Add RAG, Flows, Sparks overlay, MCP integration, built-in web search, image/video generation, and Unix commands — it's a full AI operating environment, not a pretty textarea.
Yes — and it's not basic file attachment. Qark has a full RAG pipeline: vector search, hybrid retrieval, cross-encoder reranking, HyDE generation, and code-aware chunking. Smart routing picks the right strategy per document. Drop in PDFs, DOCX, spreadsheets, code files, or entire folders. Every answer cites its source.
Yes. Sparks (Cmd+Alt+Space) is a floating overlay that works from any app — highlight text, summon Sparks, run a Flow, and the result goes straight to your clipboard. No copy-pasting between windows. Flows are reusable LLM pipelines you define once and trigger instantly.
Deeply. MCP (Model Context Protocol) lets you plug in Notion, Google Sheets, GitHub, Slack, databases, or any custom tool — agents discover and use them automatically. On top of that, Qark ships with 9 built-in tools including web search, web fetch, image and video generation, and sandboxed Unix commands.
Your purchase includes all updates and new features for 12 months. After that, the app continues to work forever — you just won't receive new features unless you optionally renew for $19/year.
Yes! Every plan includes a 7-day free trial with full access to all features. No credit card required to start.
For $99, you get lifetime updates forever — no renewal needed. This is limited to 500 licenses, then it's retired permanently.
You pay once for Qark itself. AI usage is separate — you use your own API keys and pay providers directly for tokens. Most casual users spend $2–5/month. Power users might spend more, but you're always in control with per-provider budget limits and real-time cost tracking per message.
Never. Qark has no server, no cloud, no account system. API calls go directly from your machine to the provider. Your API keys are encrypted on disk with a randomly generated encryption key unique to your installation. Conversations are stored in a local SQLite database you fully own. We couldn't read your data even if we wanted to.
Yes. Run local models via Ollama or LM Studio and Qark works with zero internet connection — no API calls, no tokens, no cost. The embedded LLM handles conversation naming and tagging locally too. Perfect for air-gapped environments, sensitive documents, or just saving money.
That depends on the provider's API terms, not Qark. Most providers (OpenAI, Anthropic, Google) do not train on API data by default. Since Qark uses API keys directly, you get the same data protection as any API user. You can also use local models for complete privacy — nothing leaves your machine.
Zero. No telemetry, no analytics, no crash reporting, no update checks you didn't ask for. Qark doesn't even have a user account system — there's nothing to phone home to. You can verify this yourself: the app makes no network requests except the ones you trigger to AI providers.
Have more questions?
Download Qark free. 7-day trial, no credit card required.