Why I built this.

Eight months ago, I realised I had 3,800 LinkedIn connections, 900 Facebook friends, thousands of WhatsApp messages, and years of calendar events. But I could not answer simple questions. When did I last see David. What did James recommend at coffee. Who introduced me to that investor in Wan Chai.

The information existed – scattered across ten apps, three devices, and two decades of digital life. But it was not connected. Not searchable. Not mine in any meaningful sense.

Every "personal AI" product I evaluated wanted my data on their servers. Upload your contacts. Connect your email. Let us read your messages so we can help you. The pitch was always the same: trust us with everything, and we will give you convenience in return.

That felt wrong. My relationships, my conversations, my medical appointments – that is the most intimate data there is. I should not have to trust a startup's privacy policy to use my own AI.

So I built it local-first. Apple Silicon made it possible – a Mac Mini runs 9-billion-parameter language models at 30 tokens per second. GDPR made the data portable – every platform is legally required to give you your data in machine-readable format. The timing was right.

What it is now

7,000+
People in the graph
2M+
Data points connected
20
Platform parsers

Ostler is a personal knowledge graph that imports your digital life – LinkedIn connections, Facebook friends, Instagram follows, WhatsApp contacts, calendar events, email patterns, browser history – and connects it all into a searchable, intelligent whole.

It runs entirely on a Mac Mini. The AI models are local. The data never leaves your house. There is no cloud server and no telemetry. $24.99 per month gives you unlimited AI – no token limits, no usage caps.

An AI assistant called Marvin answers questions about your life via iMessage, WhatsApp, and email. A personal wiki auto-generates pages for every person, organisation, and topic in your graph. A conversation processing pipeline extracts facts, relationship signals, and coaching observations from recorded conversations.

How it got here

Aug 2025 First prototype. Preference extraction from Evernote notes using local LLMs. Terrible results.
Oct 2025 Three-store architecture: Qdrant (vectors), Oxigraph (knowledge graph), Redis (cache). Started working.
Dec 2025 Marvin goes live on iMessage. First AI assistant that actually knows who my contacts are.
Jan 2026 Personal wiki compiler. Auto-generated pages for 1,820 people, cross-linked with relationship data.
Mar 2026 Identity resolver merges duplicates across platforms. WhatsApp channel added. Email triage working.
Apr 2026 Instant onboarding from macOS (Safari, iMessage, Notes, Calendar, Photos). GDPR import pipeline: 20 parsers. People graph explodes to 7,000+. 459 automated tests. First beta tester onboarding.

What comes next

Friends beta is live. The product works. A growing group of testers are importing their data and shaping what comes next. Broader access is coming.

About

Ostler is a product of Creative Machines – a Hong Kong specialist consultancy that turns emerging technology into sustainable competitive advantage. Our cross-functional team of strategists, creatives and engineers has conceived and shipped award-winning AI, AR/VR, IoT and phygital products for clients in a wide variety of verticals, including some of the world’s biggest brands. We speak fluent C-suite, drink our own champagne (we run on the very latest tech we recommend), and keep overheads low so value stays with our clients.

Ostler turns those same principles on ourselves: local-first, privacy-first, productised for individuals.

Andy Massey – Founder and CEO

Based in Hong Kong. Three decades shipping at the intersection of product, technology, and design. Leads end-to-end delivery at Creative Machines. Building Ostler because the alternative is giving your soul to the cloud and hoping for the best.

LinkedIn

If you have questions, want to test, or want to tell us we are wrong about something – hello@ostler.ai