Last verified on 26 April 2026. Pricing, features, and policies change – check the source if anything looks off.
Ostler: $49.99 + $24.99/mo – closed Mac-native system, AI runs on your Mac, data stays on your Mac.
Zo Computer: free / $18 / $64 / $200 per month – your own cloud VM with AI bundled in. Inference happens in the cloud (or via your own provider keys).
We genuinely like what Ben and the Zo team are doing. The core idea – give every user their own cloud VM that they can run anything on, instead of a locked-down chat product – is one of the more thoughtful answers to the cloud-AI control problem we have seen. They started thinking about importing your GDPR data around the same time we did last year, with a different goal (dashboards, visualisations, retrospective insight) versus ours (the AI's working memory, queried every interaction). Same instinct that your data should sit somewhere you control. Different mechanism. We respect the call.
| Ostler | Zo Computer | |
|---|---|---|
| Your data lives on | Your Mac, never leaves | Zo's cloud VM under your control |
| AI inference happens | Locally on your Mac (Ollama) | Cloud (Zo's bundled or your own provider keys) |
| Price | $49.99 once + $24.99/mo | Free / $18 / $64 / $200 per month |
| Relationship intelligence | Yes (warmth, reciprocity, history) | No |
| Personal wiki | Auto-generated, 21 page types | No |
| Multi-platform import | 20 platforms via GDPR | No GDPR import |
| Conversation capture | Meeting + messaging | No |
| Knowledge graph | Vectors + RDF triples | No |
| Web search | Yes (local via SearXNG) | Yes (cloud) |
| Works offline | Yes | No (cloud-hosted) |
| AI assistant | iMessage, WhatsApp, email | Chat interface |
Zo gives you a personal cloud computer – a real VM, not a sandboxed app, that you can install things on and use as a general-purpose environment with AI bundled in. That is a meaningful step beyond "a chat interface that owns all your data". You decide what runs there. You can swap models. You can bring your own API keys so your queries flow through providers you have an account with rather than through Zo's bundled inference. That is more control than you get from most cloud products and we like it.
Ostler takes a different route to the same instinct: instead of giving you a VM you control on someone else's hardware, we give you software that runs on hardware you already own. Same goal (you should know where your data lives). Different mechanism (your Mac, not their data centre).
Zo are upfront about the trust model in their own docs: in principle, the operators of the underlying cloud have access to the hardware your VM runs on. That is a trust statement, not a technical guarantee. It is also true of every cloud provider, so it is honest framing rather than a knock on Zo specifically.
Ostler's equivalent statement is shorter: there is no underlying cloud. The Docker containers run on your Mac, the AI model loads into your Mac's RAM, the iOS companion talks to your Mac directly over your home Wi-Fi. We do not need a trust statement because there is nothing for us to access.
Different threat models. Zo's is "the cloud provider is trustworthy and you control what runs there"; ours is "the data never leaves the device, full stop". Both legitimate, depending on what you are protecting against.
Zo offers two paths and we think the choice between them is the most interesting trade-off in their product. You can use their bundled inference (covered up to a per-tier cap on the paid plans, with usage clocking up faster than the base $18/mo plan typically covers if you are leaning on the assistant for daily tasks). Or you can bring your own API keys for OpenAI / Anthropic / etc, in which case Zo proxies your queries to providers you have an account with and you decide where the data goes. They may also run their own hosted models, which is less of a third-party-data-sharing concern than the bundled-API path but still cloud-hosted – worth checking with them directly on the current state of that.
Ostler runs the model on your Mac. Smaller model, no cloud bills, no third party in the loop, queries do not leave the device. And the inference is effectively unlimited – you can ask the assistant a thousand questions a day, every day, for the same flat monthly subscription, because the compute cost is your electricity bill rather than a metered per-query API charge. There is no token cap, no rate-limit page, no "you have reached your monthly usage" surprise. That is the trade-off in plain terms: Zo gives you frontier-grade reasoning at the cost of cloud routing and metered usage; Ostler gives you a 9-billion-parameter open-weight model that runs locally, has no usage cap, and gets better every quarter as the open-weights ecosystem ships.
Both products had the same instinct around GDPR data import last year. Zo's angle is dashboards and retrospective visualisation – bring in your data, see what your year looked like, get analytical insight. Ostler's angle is ongoing use – your data is the AI's working memory, queried every time the assistant answers a question about your life. Different products, same data source. Worth knowing if you are choosing between them.
Zo's free tier gets you a working environment. Paid tiers run $18 / $64 / $200 per month depending on compute and storage. If you bring your own keys you also pay your AI provider, on top of Zo's tier.
Ostler is $24.99/month with a one-time $49.99 base. The compute runs on hardware you already own, so the AI cost does not scale with usage. You can ask the assistant a thousand questions a day and the bill is the same as asking ten.
If you want a cloud VM you control plus an AI chat layer over it, Zo is one of the more thoughtful options in the category and worth a serious look. If you want your AI and your data on your own Mac, with no cloud routing on the inference path and no "but in principle the provider has access" caveat, Ostler is what we built. Different problems, both legitimate. We respect what Ben and the Zo team have built; if you end up using their product, that is fine by us, and we hope the local-first / cloud-with-control conversation is healthier with both of us in it.