Straight answers to what people ask most before they buy – what Ostler is, how it works, where your data lives, what it costs, and what happens if you ever walk away. Cannot find yours? A human reads every message at support.
Ostler is a private, local-first personal AI for your Mac. It reads the digital life you have already lived – your messages, meetings, notes, calendar and more – and compiles it into a searchable memory on your own machine, plus an assistant that can answer questions about it. Everything, including the AI model itself, runs on your Mac. Nothing is sent to a cloud server.
Ostler turns scattered history into one place you can actually use: a wiki of your life – people, places, conversations and topics – and an assistant that knows it. Ask “where does Priya live now?” or “what did I promise Tom?” and it answers from your own records, with the source cited. See the overview and the demo.
Anyone whose life is spread across too many apps to keep track of, and who would rather that memory live on a machine they own than on someone else’s servers. If you value privacy, dislike renting your own data back from Big Tech, or simply forget the details that matter, Ostler is built for you. Read Take it back for the thinking behind it.
No. The chat assistant is one way in, but the heart of Ostler is the compiled memory underneath it – a structured graph of your life that the assistant reads from and cites. A general chatbot knows the internet; Ostler knows you, because it is grounded in your own archives and answers from them rather than guessing.
The everyday things a good assistant with a perfect memory of your life could answer: when you last saw someone, what a friend recommended, the reply you meant to send, what you agreed in a meeting, where you read something. Because answers are drawn from your records, each one comes with its source and date. The demo shows this in action.
Three moves. First, you bring in the archives you already have. Second, a model running on your Mac reads them and works out who is who, weaving everything into a single graph. Third, out comes your second brain: a wiki of your life and an assistant that answers in the channels you already use, or on your iPhone. No cloud is involved at any step.
The same person shows up as different handles across apps – a WhatsApp number here, a LinkedIn name there. Ostler recognises that these are one person and merges them, so “Priya S” on WhatsApp and “Priya Sharma” on LinkedIn become a single entry with one shared history. That cross-referencing is what lets it answer questions no single app could.
Ostler is built to answer from your data and to say when it does not know rather than fabricate. Answers cite the source and date they came from, so you can check them. On smaller Macs the assistant is deliberately conservative and refuses to invent when it lacks the facts.
Every fact is scored and ranked, and left to fade over time the way real memory does, so what matters most rises to the top and stale detail sinks. That weighting is part of what makes the assistant’s answers feel relevant rather than a flat data dump.
No. The AI model, the databases and the whole processing pipeline run on your Mac. There is no cloud backend doing the work in the background and no per-query bill. Read why local for the reasoning.
On your Mac, and only on your Mac. Your readable data sits in plain files under ~/Documents/Ostler/, alongside the local databases that power search and the graph. There is no account, no server-side copy, and nothing on anyone else’s infrastructure. See the privacy page.
No. Ostler does not connect to any external server for your data – no cloud backend, no API endpoint, no telemetry. Personal data only ever flows inward, never out. The one caveat is that the assistant may fetch public reference material (such as Wikidata or web search results) to enrich answers; your private data is never part of that traffic. Details on the security page.
None. There is no usage tracking, no crash telemetry, and no analytics phoning home. The whole point of a local product is that there is nothing to phone home to.
Whole-disk FileVault (AES-256-XTS) protects everything on the Mac today, and the installer checks it is enabled. On top of that, Ostler is building a finer-grained, per-store encryption layer: the SQLite databases use SQLCipher (AES-256), and a per-store data-encryption key wrapped under your passkey is in review. The security page publishes an honest, component-by-component status rather than a blanket claim.
Ostler is passkey-primary. On first run it registers a passkey against your Mac’s Secure Enclave, so unlocking is a Touch ID tap, a Face ID glance, or a double-click on your Apple Watch. The cryptographic proof lives in your Mac’s hardware and cannot be exported, phished or copied. Your passkey syncs to your other Apple devices through iCloud Keychain, end-to-end encrypted by Apple.
There is no password-reset email and no magic link, because the chain of custody never runs through your inbox. If iCloud Keychain has not restored your passkey to a new device, the twelve-word recovery phrase you wrote down at setup is how you get back in – it unwraps an independent copy of the same encryption key. Keep that phrase somewhere safe and offline.
No. Every Hub service binds to 127.0.0.1 (loopback only) and no port is exposed to the internet or even the local network. If you want to reach your assistant from your phone while away from home, the recommended route is Tailscale – zero-trust, encrypted, with no exposed ports.
We are explicit about this on the security page. Because the assistant binds to your Mac, malware that already runs as you on the same Mac could reach it. Encryption at rest defends against physical theft and against your data leaving the device; it does not isolate Ostler from other software you have chosen to run. Treat your Hub Mac the way you would a password manager or banking app.
No. Your data never reaches us. There is no server-side copy to read, subpoena or breach, because your memory is compiled and kept on your own machine. That is the entire design.
Around twenty sources across your digital life. Two kinds:
The Replaces and Getting started pages list the full set.
Two ways, which run side by side. Grant Full Disk Access and Ostler reads what is already on your Mac in minutes. For the platforms that only hand data over as an export (WhatsApp, LinkedIn, Facebook and so on), you request each export once – they take anywhere from a few hours to a few days to arrive – and drop the files into Downloads for the importer. Full steps on Getting started.
It is recommended, not required. Full Disk Access is what lets you see value in minutes from data already on your Mac, rather than waiting on exports. Decline it and Ostler simply skips those instant sources; you can still import from platform exports, and you can grant access later. These are the standard Apple security prompts every Mac app must request.
Yes, if you let it. iMessage is read directly from your Mac with Full Disk Access; WhatsApp comes in through the account export you request from within the app. Both stay entirely on your machine – reading your messages is exactly how the assistant can answer questions about your conversations.
Yes, with Ostler Pro. Once set up, it keeps up with the conversations you choose – spoken, WhatsApp, iMessage, email – plus your browsing and calendar, so the graph grows richer over time. The one-time import gives you your history; Pro keeps the memory current. See pricing.
No. The installer works with whatever exports have arrived and simply skips the rest. You can run the import again later as more come in – it is repeatable – so there is no need to wait for the slowest platform before you begin.
The Ostler Hub is $99, one time. That is Ostler on your Mac, and it is yours forever. Ostler Pro is $9.99 per month, with the first 30 days free, and adds the iPhone app plus the ongoing ingestion that keeps your graph current. Full breakdown on the pricing page.
Hub buys you Ostler on your Mac forever – the wiki, the timeline, the people graph, the Mac app, and the assistant answering locally over your snapshot, including via iMessage and WhatsApp. Pro extends Ostler to your iPhone and keeps ingesting new data into your graph. Hub is the foundation; Pro keeps it alive and puts it in your pocket.
Yes. $99 once. Because the model runs on your own hardware, there is no cloud bill and no token limit – the more memory your Mac has, the more capable it gets, at no extra cost. Your first 30 days of Pro are included with every Hub purchase.
The clock starts when you start Pro, not when you buy the Hub, so you can redeem it whenever you are ready. After the free month it is $9.99 per month via the iOS App Store, and you can cancel any time.
No. No cloud bill, no token limits, no metered usage. The Hub is a single purchase and Pro is a flat monthly subscription. Everything the assistant does runs on hardware you already own.
Ostler Pro is billed by Apple, so you cancel any time in your Apple ID subscription settings. Your existing data stays on your Mac, exactly as you left it – see what happens to my data below.
An Apple Silicon Mac running a recent version of macOS – any of the Mac Mini, Mac Studio, iMac, MacBook Air or MacBook Pro line. One licence runs on any of them. For Ostler Pro you also need an iPhone for the companion app.
16 GB is the floor and runs the compact assistant – reliable, accurate and sub-second on short questions. 24 GB is recommended for a richer answer range and is what the team uses daily. 8 GB is not enough.
Yes. Your hardware is your AI: the more memory your Mac has, the larger the local model it can run and the richer its answers. 48 to 64 GB unlocks bigger models with headroom to grow; 96 to 128 GB runs frontier-class local models – about as good as it gets without the cloud. See the hardware guide on the pricing page.
Ostler is built for Apple Silicon Macs. It relies on the on-device performance and the Secure Enclave that Apple Silicon provides, so an Intel Mac or a Windows PC is not supported.
Not for the Hub. The Mac does all the thinking and the assistant is reachable on the Mac and via iMessage and WhatsApp. You need an iPhone only for Ostler Pro’s companion app, which puts your assistant in your pocket and pairs with the Hub through your passkey.
About 30 minutes of hands-on time across three steps. The one part you cannot rush is the platform exports (LinkedIn, WhatsApp and the rest), which take a few hours to a few days to arrive – but you request those once and can start importing whatever has landed. Walk-through on Getting started.
You drag in your licence file, and the installer checks your hardware, sets up the local model and databases, and registers the Hub services with macOS. It then walks you through security setup: registering your passkey and saving your recovery phrase. It also guides you through a short set of setup questions to get your graph started.
Within minutes if you grant Full Disk Access, because Ostler can read data already on your Mac straight away rather than waiting on exports. The first read-in then continues in the background, and the wiki fills in and sharpens over the next few days as the feeds finish settling.
Yes. After the first read-in, cards and pages keep filling in over the following days, and with Ostler Pro the graph grows richer every week as new conversations, browsing and calendar activity flow in. It settles into itself rather than arriving fully formed on day one.
You need the Hub on a Mac – that is where Ostler lives and thinks. The iPhone app comes with Ostler Pro and is optional, but it is how you reach your assistant anywhere and keep the graph current. Many people run the Hub on an always-on Mac at home and carry the Pro app on their phone.
Whichever suits how you live. An always-on Mac Mini, Studio or iMac on a desk means Ostler is ready the second you are and your iPhone can always reach it. A MacBook works too: when the lid is closed, captures queue locally and the assistant catches up the moment you open it again. One licence – your call which Mac wears the Hub hat.
Yes, with Ostler Pro. Because the Hub binds only to your Mac and exposes no public ports, remote access goes over Tailscale – encrypted and zero-trust – so your phone reaches your own Hub without anything being open to the internet.
The Hub runs on your Mac, so it is offline whenever the Mac is asleep, shut down or off the network. Wake it and the assistant reconnects automatically. The iOS app shows a clear Online / Catching up / Offline pill so you always know whether an answer is live or stale.
The Hub stays yours forever. Your wiki, timeline, people graph, the Mac app and the assistant answering locally over your snapshot all keep working whether or not Pro is active. What pauses is the ongoing intelligence: new messages, emails, browsing and photos stop flowing into the graph until you restart Pro. Everything you already have is untouched, and if you restart, it picks up where it left off.
Ostler is designed not to punish you for a card decline, a brief pause, or a holiday with no signal. Your Hub keeps working locally either way, and if the Hub cannot temporarily confirm your Pro status, the Doctor tells you so and points you at a fix. Restart Pro at any time and ingestion resumes from where it left off – your graph is where you left it.
No. The uninstaller removes the engine under ~/.ostler/ and stops the local services, but it leaves your visible data at ~/Documents/Ostler/ intact – your conversations, wiki pages and exports stay put. If you also want the visible data gone, you delete that folder yourself. Steps are on support and in the docs.
Your data lives in plain Markdown under ~/Documents/Ostler/, alongside standard local databases (Qdrant JSON, Oxigraph Turtle / N-Triples / JSON-LD, and SQLite). It is all open, standard and readable – you can open, search, correct or delete it yourself. Subscription state never gates your data. As Take it back puts it, there is no export button between you and your own past, because it was never anywhere else to begin with.
Yes. It is plain files on a computer you own. Nothing is held hostage on a server, there is no account to close to “free” your data, and cancelling or uninstalling never reaches back and deletes what is on your Mac. You can pull the plug at any moment and keep everything. More on It’s yours.
The assistant answers from your local graph and local model, so the core product works without the cloud. An internet connection is needed for a few normal things – requesting your platform exports, downloading updates, reaching your Hub remotely over Tailscale, and optionally pulling public reference data – but answering questions about your life does not depend on it.
A capable open-weight model runs locally on your Mac via Ollama – for example, a Mac Mini M4 comfortably runs a 9B-parameter model at conversational speed. It is not a frontier cloud model, and it does not need to be: the task is your personal knowledge, not general intelligence. The bigger your Mac, the bigger the local model it can run. More on why local.
Your personal data only flows inward, never outward. Ostler may fetch public data from the internet – things like Wikidata or web search results – to enrich an answer, but your private data is never sent out as part of that. There is no cloud endpoint receiving your messages, notes or graph.
No. Your data never leaves your Mac, so there is nothing for us or anyone else to train on. The local model reads your archives only to build your private graph; it does not upload them or fold them into a shared model. This is the opposite of the cloud-proxy approach, where your data passes through someone else’s servers.
ChatGPT is a brilliant generalist that knows the internet but not your life, and runs in someone else’s cloud. Ostler knows you – it is grounded in your own archives, cites its sources, and runs entirely on your Mac with nothing sent out. They solve different problems. See the full comparison.
Both build a memory of you, but the approach differs. Rather than continuously recording your screen, Ostler compiles the archives you already have – messages, notes, calendar and the rest – into a structured graph on your own machine. We lay it out side by side on Ostler vs Rewind.
A notes app stores what you type; Ostler compiles what you have already lived across twenty-odd sources and cross-references it into one memory, then puts an assistant on top that can actually answer from it. It even writes to plain Markdown, so it plays nicely with tools like Obsidian. See Ostler vs Obsidian.
Apple’s on-device features are a strong signal that private, personal AI is the future, but they are scoped to Apple’s own apps and moment-to-moment tasks. Ostler cross-references your whole life across many platforms into one durable memory you own and can inspect. We cover it on Ostler vs Apple / Siri.
The Compare page puts Ostler side by side with the major alternatives on privacy, price, features and data location, and links to a detailed page for each one.
Email [email protected] and a human replies, usually within one business day. Most operational how-tos live on docs.ostler.ai, and the support page answers the most common issues.
The Hub is non-refundable once delivered, because the import runs immediately on purchase and writes your data to your Mac in open formats – at that point you have already received the product. EU and UK buyers tick a box at checkout consenting to immediate delivery and waiving the 14-day withdrawal right; statutory rights for non-conforming content are unaffected. Ostler Pro is billed by Apple, so Pro refunds follow the App Store refund policy. Full detail in the Terms of service.
Please use the disclosure path on the security page rather than emailing support. Reports go to [email protected], a human reads every one and replies within 72 hours, and there is a published PGP key at /security.asc and a machine-readable policy at /security.txt.
At docs.ostler.ai – install, recovery key, the Doctor panel, the conversation pipelines and the wiki are all covered there in depth.
Still have a question? Write to [email protected] and a human will answer. If it turns out to be a question other people are asking too, it will end up on this page.