Building Mythra: a desktop AI app with personal Wizards and collaborative projects
For the past few months, I’ve been building an app called Mythra.
The goal was to create a desktop application for both macOS and Windows where people could use any AI model they wanted in a clean, modern interface.
At its core, Mythra is a chat-first desktop app. You can connect your own models through OpenRouter or run local models directly. The interface should feel familiar to anyone who’s used contemporary AI tools, but the focus is on speed and clarity rather than flashy distractions.
From chat app to agents
What started as a simple chat app evolved once I began experimenting with giving the models more autonomy.
I added the ability for AI models to act as agents that can perform actions on your behalf inside the app. These agents can handle tasks, maintain context across sessions, and actually do work instead of just responding to one-off prompts.
That shift changed Mythra from something I could talk to into something I could actually work with.
Wizards
My favorite feature ended up being something I call Wizards.
They work similarly to the kind of persistent agents people have experimented with elsewhere, but with a much cleaner and easier-to-manage interface. You can create, personalize, and steer individual Wizards over time. Each one remembers who you are and what you care about, rather than treating every conversation like it’s meeting you for the first time.
That persistence is what makes them feel useful to me. They aren’t just temporary chats with a personality layer on top; they feel more like long-term assistants you can actually shape.
Nexus
The feature I’m most excited about is Nexus.
With Nexus, you can create projects and assign multiple Wizards to work on them together. These Wizards can communicate with each other, share context, and coordinate on how to best accomplish the goals you set.
That turns Mythra from a single-agent tool into something more like a small team you can direct.
Why I’m building it
I built Mythra because I wanted an AI environment that felt truly mine.
One where I could choose the models, shape the agents, and organize work the way I actually think about it.
I’m getting close to sharing it more widely, and I’ll have more details soon.