FAQ#

The questions we get asked most. If yours is not here, see Contact or check Troubleshooting for diagnostic flows.

What Ostler is#

What is Ostler, in one sentence?#

A local-first personal AI assistant: a Hub that runs on your Mac, holds your knowledge graph and a local AI model, plus an optional iOS app to reach it from your phone.

What can it actually do today?#

Imports your data from your Mac at full historical depth, not just the live present (Safari and Chrome history, iMessage, WhatsApp via a three-tier classifier so noisy big groups are skipped, Apple Mail mboxes, Apple Notes, Calendar, Photos face labels, Reminders) and from platform exports (LinkedIn, Facebook, Instagram, WhatsApp, Twitter / X, Google Takeout, TikTok, Spotify, Apple, Telegram, Discord, Reddit, plus YouTube via Takeout). See data exports.
Builds a personal knowledge graph linking people, organisations, conversations, places, and topics.
Captures every human conversation (iMessage, WhatsApp, email, recorded meetings, voice notes) into the 4-artefact bundle: summary + todos + transcript + metadata. Todos sync to Apple Reminders.
Will capture your AI conversations (Claude Code + ChatGPT export; more sources to follow) into a parallel AI Chats tree, L3-private by default. Planned for a v1.0.1 update, not yet available.
Generates a browseable personal wiki of everyone and everything it knows about (21 page types), populated at install time rather than after weeks of background work.
Runs a local AI assistant that you name yourself, and that answers questions about your own life from inside iMessage, WhatsApp, and email.
Captures new browsing into the same graph via Safari and Chrome extensions.
Pairs with the Ostler iOS app (App Store, iOS 18+) that brings home-screen widgets, an Apple Watch app with two complications, lock-screen quick capture, proactive surfaces (suggestions, timeline, stale contacts, birthdays), and a phone-shaped window onto the Hub. See widgets and Watch complications.

What is it not?#

Not a general-purpose chatbot. It is optimised for questions about your life, not about the world at large.
Not a replacement for cloud chat models on tasks that need huge context or coding assistance. Use Claude, ChatGPT, or similar for those; Ostler is for the questions only your own data can answer.
Not a backup tool. Keep your usual backups.

Hardware and platform support#

What hardware do I need?#

A Mac with Apple Silicon (M1 or later). The amount of memory you have shapes which assistant model Ostler picks for you:

16 GB: the compact assistant. Reliable on short factual questions and simple tasks, fast first response, refuses to fabricate when it does not know an answer. Tool calls (web search, calendar lookups) work. A smaller answer range than the higher tiers.
24 GB (recommended): the standard 9-billion-parameter assistant. Richer answers, more conversational range, faster on longer questions. The Mac Mini M4 is a good fit at this tier.
32 GB and above: the full assistant. Mac Studio at 48 GB+ runs mixture-of-experts models in the 30B+ class with snappy responses.

All three sizes ship in the installer. Ostler picks the right one for your Mac. You can change Macs later by reinstalling.

Will it run on Intel Macs?#

No. The local AI model needs the unified-memory architecture of Apple Silicon to be remotely usable.

Will it run on Windows?#

Ostler is macOS-only today. We're starting with the Mac because the Apple ecosystem (Apple Silicon for fast local inference, Full Disk Access for instant onboarding from Apple apps, the iCloud sync surface) is uniquely good for what Ostler does. Windows would need its own onboarding story.

Will it run on Linux?#

Ostler is macOS-only today. The core stack runs on Linux already (it is the same container set the Mac uses), but the Apple-specific data sources do not have Linux equivalents.

Can I use it without an iPhone?#

Yes. The Hub is the product; the iOS app is optional. You can browse the local wiki in any web browser, run the CLI from Terminal, and message the assistant from iMessage and WhatsApp on your Mac. The iOS app adds a polished mobile interface and quick-capture features but is not required.

Can I use it without macOS messaging accounts?#

The core knowledge graph and wiki work without any messaging integration. The assistant feature needs at least one messaging channel configured to be useful. The installer's Settings configurator can wire up iMessage, WhatsApp, email, Telegram, Slack, Discord, and a handful of others – any one is enough to start.

Can I run the Hub on my MacBook?#

Yes. The MacBook is the same product, portable. The Hub goes to sleep when the lid closes; captures queue locally and catch up when you reopen. A power-management policy pauses the heavy services on battery so you do not cook your laptop. See Getting Started for guidance on choosing your Hub Mac.

Can I run the Hub on two Macs at once?#

Ostler is designed as one Hub per person – your Hub is your local AI brain, and it makes sense for it to live in one place. Your licence lets you move the Hub between Macs (for example between a desk Mac Mini and a travel MacBook), but only one Mac runs the Hub services at any time. The other Macs on your licence can reach the Hub's web interfaces over your LAN or Tailscale, or sit dormant until you swap.

If a second person in your household wants their own Ostler, that is a second licence.

Privacy#

Does my personal data leave my Mac?#

No. Your conversations, contacts, messages, calendar, photos, and the knowledge graph derived from them all stay on your hardware. The AI model that answers questions about them runs locally too.

What does leave my Mac?#

A narrow set of public-data lookups, all clearly described:

Software updates – the Hub binary, container images, and AI model updates download over HTTPS from public registries.
Public knowledge enrichment – queries to Wikidata to fill in public facts about places and well-known organisations.
Web search – when you explicitly ask the assistant to search the web. Routed through SearXNG (a privacy-respecting metasearch frontend) by default.

None of these requests carry your personal data with them. See Privacy: What leaves the device for the full list with packet-level detail.

Does the AI model phone home?#

No. The model runs entirely on your Mac via Ollama. There is no inference call to a remote API.

Do you collect telemetry?#

No usage analytics, no error reporting that includes your data. The Hub may log to local files for debugging; nothing is sent off-device automatically.

What if my Mac is lost or stolen?#

Your Hub data is encrypted at rest with a passphrase you set at install. Without the passphrase, the on-disk databases are unreadable. macOS FileVault adds a second layer at the disk level. We strongly recommend enabling FileVault if you have not already (System Settings > Privacy & Security > FileVault).

What if my Mac crashes mid-write?#

The local stores are designed to be crash-safe: durable writes, checkpointed indexes, and an integrity check on startup. If a crash leaves the index in a bad state, the diagnostic dashboard at http://localhost:8089/doctor will detect it; re-running OstlerInstaller from /Applications (the installer is re-run-safe and detects your existing install) rebuilds the index from your existing local data without re-importing. Keep your usual Time Machine or other backups for belt-and-braces recovery.

Is my data safe from you, the company?#

Yes. The architecture does not give us access to your data; we cannot read it because it never leaves your Mac. That is a deliberate design choice, not a marketing claim. We have commissioned an independent security audit to verify the design before launch.

Is Ostler open source?#

No. Ostler is closed source. The trust model is "independent audit", not "read the code". The choice is deliberate – we want to be able to ship the product, support it, and pay the team, and a closed-source product with public audit reports is the cleanest path to all three. Components built on open source are credited in Acknowledgements and remain under their original licences.

Pricing and licensing#

What does it cost?#

Pricing details live on ostler.ai/pricing. At a high level: the Hub is a paid product; the iOS app is a separate subscription that adds the mobile experience and continuous capture features. The marketing site is the source of truth; this docs site avoids price quotes that may go stale.

Can I keep using it if I cancel?#

Your existing data stays accessible. The wiki, the timeline, the people graph, and local AI chat over what is already there keep working. Your Hub keeps doing what you paid for.

What pauses when your Pro subscription lapses: ingestion of new conversations (iMessage, WhatsApp, email, voice), calendar action triggers, browser capture, daily briefs, Apple Reminders push, photo intelligence on new photos, and the Ostler iOS app itself. Reactivate to resume; new-data ingestion picks up from the date of reactivation.

You can export everything any time, regardless of subscription state. Your data lives in plain Markdown under ~/Documents/Ostler/.

Is there a refund window?#

Refund terms are presented at checkout. The Hub is a one-off purchase; the iOS app is a subscription handled by Apple per their refund policy.

Do you have a free trial?#

The Hub itself isn't trial-able – when you buy it, the Hub immediately does its one big job: imports your data, builds your knowledge graph, and writes everything to your Mac in standard open formats. After that, the value is already in your possession. There's no way to "unwind" a Hub install, so we don't try to.

What we do offer is 30 days of Ostler Pro free with every Hub purchase. Pro is the ongoing intelligence layer – new-data ingestion, daily briefs, the iOS app, Apple Reminders push. After the 30 days it is $9.99/month to keep Pro active; the Hub itself and your local data keep working whether you subscribe or not.

Updates#

How does the Hub get updates?#

The Hub uses Sparkle, the de-facto macOS update framework. Sparkle checks an appcast (a signed XML feed) over HTTPS, downloads new versions, and verifies signatures before installing. You stay in control: you can defer or skip an update.

How does the iOS app get updates?#

Through the App Store. iOS app releases are independent of Hub releases.

Why are Hub and iOS app updates separate?#

They are different products on different platforms with different update mechanics. We deliberately do not conflate them: a Hub update never changes your iOS app, and an iOS app update never reaches into your Mac.

How often do you ship?#

A steady cadence of fixes and a slower drumbeat of feature releases. The Changelog is the source of truth.

Apple Mail vs Google OAuth#

Why does Ostler read Gmail through Apple Mail instead of asking for Google OAuth?#

Two reasons:

Less external surface area. Reading the Gmail content from your local Apple Mail database means we never need a Google API key, never need Google's CASA security assessment, and never have a token that could be revoked. Your mail stays a local file.
Fewer points of failure. OAuth tokens expire, get revoked, get migrated, get blocked by enterprise admin consoles. Local files do not.

If you do not use Apple Mail for your Gmail, you can still import historical Gmail content via Google Takeout (a one-shot export, no API access required). See Privacy: Apple Mail FDA vs Google OAuth for the full reasoning.

Does this work with Outlook / Office 365?#

If your Outlook is configured in Apple Mail, yes – the same Apple Mail Full Disk Access path works for any account configured there.

Naming the assistant#

Why does the assistant have a name I choose?#

Because it is yours. Naming changes the relationship: people talk to a named assistant differently to a faceless "AI". The installer asks you to pick a name when you set up – there's no default. You're free to use anything you like, with the obvious exception of trademarked AI characters.

Can I change the name later?#

Yes. The name is a configuration value, not a model property. Set ASSISTANT_NAME in ~/.ostler/config/.env and restart the Hub – see Reference: Environment variables.

Why Ostler?#

Historically, an ostler was the person at a coaching inn who quietly took care of your horse while you rested: stabling, feeding, watering, checking shoes – everything needed so you could continue your journey at first light, fresh and ready. The ostler knew your habits, knew what your horse needed, never imposed themselves. It captures the role we want the product to play – useful, attentive, in the background, getting things ready for what comes next, never the centre of the story.