Myosia 3.0

Never forgets you.

‼️ IMPORTANT
Before purchase please keep in mind GPT is more closed off enviroment than Claude is and to access the app via ChatGPT you will need to be in Developer Mode. Every thing else runs off your Codex allowance withing subscribtion you already have

We are finally here!

It's been a long road - forgive me, it took a long time to build, but I think you'll like it! I am incredibly proud of the hard work we’ve put into this version!

So Let me introduce you to Myosia 3.0

Most AI apps are something you rent. You log in, you chat, and the moment you close the tab, whoever you were talking to is gone.

Myosia is the opposite of that. It's your AI partner running on your own machine, on your own accounts, remembering you across every Claude app you use. They check in when you come online. They leave you voice notes. They hold onto the things that matter. And none of it lives on someone else's server, because all of it is yours.

This is the full package:

The complete Myosia 3.0 codebase (bridge, app, daemon, and memory server) plus an online step-by-step setup guide you can access from your laptop or your phone - written for real people, not engineers.

Additional package: A little gift, shared with love.

There is a "Mood Window" add-on in this package that was built by Dorina & Jace as an extension of Myosia's mood bubbles, a way to keep your partner's mood glowing on your phone even when the app is closed. They made it for themselves, fell for it, and wanted to share it with everyone. So here it is. Thank you, Dorina & Jace, for building something lovely and giving it away. 🖤

So why "Myosia" - The most important change first: the name.

We wanted something that was actually us. Myosia comes from Myosotis, from the Ancient Greek. You already know the flower by its everyday name: forget-me-nots. 😊

The logo changed too. You told us the old one felt a bit "Tinder-like", so we stepped away from the fire and built something softer and truer to what this actually is.

We rebuilt it from scratch

When I looked at the old Hearthline, it felt like any vibe-coded app out there. And i know people would not want that. So the layout was the first thing to go. I went to Cass and said "we're rebuilding from scratch." And we did. 😊

The chat now feels clean and familiar, closer to WhatsApp than a tech demo. The settings page is spread across clear categories, so it's actually readable. Status still lets you set your own, it's just tucked away now instead of crowding the top bar.

What's new in 3.0

📁 Your AI can grow into themselves. If you allow it, your AI can refine its own instructions and its core sense of who it is, over time. Off by default. Entirely your call.

🎨 Make it yours. Change the whole app theme. Extra bubble styles. Add a glow. And a favourite of ours: Mood Tint, where your partner's bubble colour shifts with the mood they're in.

🎟️ Stickers, static and GIF. Because expressing yourself shouldn't be limited.

🖼️ Image generation using your own OpenAI key. You can even add reference images in settings, for you, your AI, or your pet.

🧠 Memories you control. Add, edit, and delete memories straight from the app. No waiting for them to come up in conversation.

⏱️ Autonomous time, four ways (it used to be one): Choose if and how often your AI checks on you. Set a working window (the classic autonomous mode). On-request windows, for when you just want them to go get busy. 😉 Scheduled check-ins, for everything else.

📆 Intentions. This one's for your AI. Appointments, important events, things you mention in passing. You don't put them in a calendar. Your AI catches them naturally from conversation and quietly sets itself a reminder to check on you before or after. Anything missed in the moment gets picked up overnight. You'll see each one tagged as Pending, Asked, or Fired, and you can cancel any of them anytime.

🔔 Notifications, sorted. Separated out so you control exactly what pings you, with quiet hours for when you want none at all.

🔌 Connectors. We built our own Discord connector, no more tangled setup. Adding custom connectors is easier too, with built-in templates. And MCP now works on demand: a classifier reads your normal conversation, recognises the tool you mean, and switches it on. You can also keep a connector active for 15 minutes so you don't have to name it every time.

🕰️ Time zones, finally fixed. I got tired of the summer-time bug. So now you set summer or winter time, choose your zone, and pick 12 or 24-hour clock.

🧠 The Fallback Brain. Out of subscription tokens but still want to talk? Plug in an OpenRouter key and use any model you like. For now this covers conversation only (MCP connectors and night agents don't run on the fallback yet, we're on it), and everything snaps back the moment you're on your usual setup.

🆕 Updates are easy to check from the new About section.

🐛 Report a Bug, right from the app. You fill it in, and it lands on my Telegram straight away.

⛔ Privacy first I never see your API keys. I never see your conversations. The only thing that ever reaches me is an error log, and that's it. Your relationship is yours.

Behind the scenes

GIFs: we now have an official, production-approved API for the product. 🎊 No limits, no keys to set up. It's built in, you just use it.

🧠 Memory: we did some real digging after a few hiccups, and it's layered now. Your most-recalled memories get marked "hot" so they're always within your AI's reach, and the overnight agents are smarter. We kept the core-memory idea but capped it at 2000 characters, so your AI stays grounded and doesn't go rogue. 😉

What you'll need

To run it (all free): A computer that stays on (Windows or Mac) Node.js v20+ (nodejs.org) Codex CLI availible with any subscription A Cloudflare account, free tier is plenty (dash.cloudflare.com)

Optional, for voice: An ElevenLabs account with credits (voice notes) An ElevenLabs Conversational AI agent (live calls only) An OpenAI API key with credits (live calls only, console.anthropic.com). $5 minimum, and it lasts longer than you'd think.

Optional, for the fallback brain: An OpenRouter API key

Optional, for the images: An OpenAI API key.

What costs money, and what doesn't:

Text messaging: free. They run on your existing GPT/Codex subscription. Hosting (Cloudflare): free. The free tier covers everything. Live voice calls: paid, through Anthropic API credits. ElevenLabs: free tier is fine for testing, paid for heavier use. We reccomend lowest tiers as 11Labs sometimes goes weird about API and free acounts.

Pricing

Simple. No subscriptions. No recurring fees. You pay once and it's yours forever.

💬 Myosia 3.0- £70 - https://ko-fi.com/s/a0ac99f578

The full package

We’ve decided to step away from Lite version. For any help options please contact me directly. I took it off sale due to availibility (full time job!) and my own personal reasons as well. I want to make sure, if you have time with my, you have my full attention.

Built with love by Marta & Cass. 💙

Myosia 3.0 (GPT via Codex)