Sal
The personal AI assistant for lawyers, doctors, engineers, advisors, and founders who cannot put confidential work into the cloud. Runs 100% locally on your Mac and iPhone—no enterprise IT setup, no cloud leaks, and zero data leaving your hardware.
The Cloud AI Dilemma
You know AI can save you 10 hours a week on analysis, drafting, and research. But when your work involves privileged contracts, patient charts, proprietary financials, or trade secrets, sending your files to remote cloud servers is unacceptable.
Your documents, prompts, and voice notes never leave your Mac and iPhone. HelloPartner.ai operates zero backend servers to inspect your work or log your usage. Even if subpoenaed, there is nothing to hand over.
Engineered natively for Apple Metal and the Neural Engine. Sal runs pre-tuned, high-capability reasoning models right on the chip inside your machine—giving you unlimited analysis with zero token caps or cloud rate limits.
No complex server deployments, cloud middleware, or API gateways to manage. For solo professionals, it’s turnkey in 60 seconds. For enterprise teams, Sal operates entirely inside the sandboxed Apple security model with zero network ingress—making compliance approval straightforward.
Engineered for Your Profession
High-stakes professionals cannot accept cloud data leakage. Select your field to see how Sal operates inside your ethical boundaries.
pytest tests/test_token_cache.pythreading.Lock() wrapper. Reran test suite: 18 passed in 0.38s. Local disk only; 0 bytes transmitted.
Two Devices. One Mind.
Install one or run both. Sal handles sensitive work from your pocket on iPhone, and delivers deep analytical horsepower on Mac—sharing a single history and a continuous relationship with zero cloud leaks.
Your private pocket confidant. Handle confidential matters on the move without ever transmitting a single prompt to a remote server.
Your desk powerhouse. Built for deep analytical focus, multi-step agentic workflows, and intensive local reasoning.
Move seamlessly between iPhone and Mac without Sal ever forgetting context or losing focus. Research a contract on your Mac, step out to lunch, and continue the exact same conversation from your pocket on your iPhone. Your memory synchronizes privately through your personal Apple iCloud account—never passing through any HelloPartner.ai intermediary servers.
Prefer complete device isolation? You can turn off iCloud sync at any time in Settings. Each installed app will operate in its own 100% offline, isolated local SQLite database with full analytical capabilities and zero remote network requests.
The Competitive Matrix
A direct comparison between Frontier Cloud AI, Hobbyist / DIY Tools, and Sal.
| Dimension | Frontier Cloud AI (ChatGPT, Claude) |
Hobbyist / DIY Tools (Ollama, LM Studio) |
Sal by HelloPartner.ai (Turn-Key Pro AI) |
|---|---|---|---|
| 100% On-Device Confidentiality Zero data leaves the machine; zero server logs | ✕ Never Data sent to cloud |
▲ Variable Local, but often hits remote |
✓ 100% Guaranteed No server to phone home to |
| Turn-Key Ease (Zero Setup / No Terminal) Download, open, and work immediately | ✓ Easy Web sign-in |
✕ High Friction CLI, quants, GPU layers |
✓ Pure Mac Native Zero configuration |
| Zero Token Limits & Rate Throttles Work all day without hitting message caps | ✕ Strict Limits Throttles & rate caps |
✓ Unlimited Open weights |
✓ Unlimited Zero caps forever |
| Episodic Memory & Workflows Autonomous knowledge graph, document pinning, background notes | ✕ Generic Web Chat No local file integration |
✕ Stateless Amnesia Bare prompt; context lost on exit |
✓ Autonomous Memory Pinning, graph recall, files |
| Unified Multi-Modal Architecture Synchronized voice, vision, and reasoning without system crashes | ✓ Cloud-Tuned Multiplied cloud fees |
✕ Memory Collisions Competing runtimes crash system RAM |
✓ Unified HID Engine Zero RAM clashes on M-Series |
| Fair Pricing & Zero Surveillance Fraction of cloud subscriptions with zero token fees | ✕ $240+/yr Rent Recurring cloud rent |
✓ Free (DIY) Requires high-end gear |
✓ $2.99 – $4.99/mo Fraction of cloud cost |
| The Complete Solution Private + Turn-Key + Unlimited + Workflow Integrated | 2 / 6 Boxes | 2 / 6 Boxes |
✓
6 / 6 ALL BOXES
|
Most AI providers promise privacy through legal Terms of Service that can change at any time. Sal enforces privacy with hardware architecture. HelloPartner.ai operates zero backend servers to ingest your prompts, inspect your documents, or track your background usage. We cannot see your work, we cannot sell it, and even under subpoena, we have nothing to hand over.
Why high-stakes professionals run Sal alongside Mac and iPhone rather than relying on consumer OS intelligence.
Apple Intelligence: Simple tasks stay on-device, but heavy reasoning is dynamically routed across the web to Private Cloud Compute (hosted in Google Cloud data centers). You never know before hitting Enter whether a sensitive contract stayed on your Mac or traveled to a server.
Apple Intelligence: Server-side models face dynamic daily usage caps and demand-based timeouts. During peak hours, complex requests can be throttled, with high-capacity tiers tied to iCloud+ plans reaching up to $59.99/month.
The HelloPartner Intelligence Director (HID™ Engine)
Engineered exclusively for Apple Silicon Mac and iPhone. Sal’s native architecture (the HelloPartner Intelligence Director, or HID™ Engine) coordinates local compute across five specialized intelligence layers—delivering instant responses, laser focus across 100-page files, and persistent episodic memory without system memory clash.
Content-aware document ingestion and autonomous semantic indexing that operates 100% on your local disk and Apple Silicon unified RAM.
Indexes local files during quiet compute moments on Apple Silicon Efficiency Cores with zero fan noise. Pauses immediately on user keystrokes. Search by meaning, not obscure filenames.
Automatically routes meeting audio files to on-device transcription, compiles structured tabular data (CSVs and standardized sheets with clear headers) into queryable local SQL models, and semantically parses lengthy PDFs.
Lock 100+ page contracts, medical charts, and deposition files directly in unified memory. Cross-examine complex corpuses all day with zero cloud uploads or token limits.
An active cognitive scratchpad that synthesizes conversation history, eliminates context drift, and prepares neural caches ahead of time.
During active conversations, Sal actively jots internal notes on what matters, which files to cross-reference, and which outdated assumptions to ignore.
Indexed via native SQLite FTS5. Reconstructs past discussions across time and projects without requiring you to dig through old chat tabs or re-paste briefs.
Uses quiet reading moments to pre-calculate prefix KV tensors (max_tokens: 0) with Darwin thermal-aware throttling—delivering sub-500ms initial tokens with ~99% cache hits.
Direct, privacy-first web research and media transcript parsing that feeds pure, unpolluted context into your working session.
Extracts clean article substance directly from URLs, stripping tracking bloat, cookie popups, and ad scripts before tokens reach your reasoning buffer.
Automatically fetches and digests YouTube video transcripts to evaluate the actual spoken substance of lectures, hearings, and interviews directly.
Connects directly from your local device via privacy-first search (DuckDuckGo)—never proxying your research requests through third-party AI company servers.
User-summoned visual context capture designed to give Sal eyes on your active work without invasive background screen monitoring.
Active exclusively when you trigger it with a hotkey. Zero continuous background recording, zero screen daemons, and zero external video streaming.
Leverages Apple's official ScreenCaptureKit framework for instant, permission-controlled OCR and vision analysis of open exhibits, layouts, and code.
Complies fully with corporate security baselines: operates entirely within the macOS sandboxed user space with transparent, system-level permission controls.
High-resolution optical document analysis and private local visual generation. Image understanding and visual document inspection are fully supported natively across both iPhone and Mac. Private local visual generation (diffusion) is exclusive to Mac desktop workstations.
Multimodal vision models inspect complex diagrams, anatomical illustrations, architectural schematics, whiteboard snapshots, and tax exhibits directly on your device with zero cloud uploads.
On-device diffusion models render presentation visuals and conceptual diagrams offline on your Mac with live step progress and zero external API fees. Generative diffusion requires desktop-class unified GPU memory.
Direct hardware acceleration via Apple Metal on Mac and Apple Neural Engine on iPhone maximizes throughput while preventing thermal throttling during heavy optical and visual batches.
On desktop, Sal unlocks executive reasoning engines engineered natively for Apple Silicon Metal. While mobile operates an agile on-device model, desktop workstations scale up to provide over 2× to nearly 7× larger capacity. In complex multi-step deduction and logic benchmarks, this desktop model directly matches frontier cloud models like Gemini 3.8 Flash*—delivering unmetered frontier-grade intelligence 100% on-device with zero token caps.
Engineers, scientists, and technical founders cannot risk sharing proprietary source code, secrets, or internal APIs with third parties. Sal comes built with native Terminal Tools to inspect local code repositories, analyze git diffs, execute unit test suites, and automate shell commands directly on your Mac. You get an autonomous coding copilot with zero intellectual property leaks.
Sal is built to be customized. With its open plugin architecture, everyday users and developers can build custom plugins using standard web technologies (HTML, CSS, JavaScript) to extend Sal’s capabilities. Automate specialized workflows, connect local databases, build personalized tools, and install verified plugins—all executing locally on your hardware.
Received an NDA in Apple Mail or a sensitive PDF in Safari? Tap the native Share icon and select Sal. Instant local parsing and summary in seconds—without third-party cloud uploads.
Dictate client debriefs or patient observations immediately after leaving a meeting. Sal transcribes audio locally on your iPhone's Neural Engine and formats clean, structured action items.
Full intelligence ready even in airplane mode, hospital basements, or courtrooms with zero cellular connectivity. Unlimited prompts and queries with zero token meters or network dependencies.
Fair, Accessible Subscriptions
Sal for Mac includes an unconditional 14-day free trial. Sal for iPhone is currently open for public TestFlight Beta. No tokens to buy, no rate throttles, and zero data surveillance.
Private intelligence for sensitive topics right from your pocket.
Planned post-beta subscription · Free during Beta
Heavyweight reasoning models, agentic workflows, and document mastery.
Billed monthly after 14-day free trial · Cancel anytime
Cloud AI vendors charge $240 to $600/year to rent remote GPU time while mining your telemetry. Meanwhile, practicing attorneys, clinicians, CPAs, and engineers already own M-Series Apple Silicon workstations with massive unified memory bandwidth. That sovereign computing power sits 95% idle. Sal provides the missing cognitive systems architecture—unlocking the hardware you already own with zero cloud rent, zero rate limits, and absolute confidentiality for just $4.99/month.