Hey Indie Hackers,
Let's be brutally honest. The AI tool situation right now is an absolute joke.
You've got Claude desktop, right? Yeah, it's got some decent ideas – "Artifacts," "MCP," "Projects." Fine, whatever. They're not terrible features. They actually solve some real pain points. They make you productive.
But then they pull the classic vendor lock-in crap. Oh, you like our features? Great! Now you're stuck with our models. And if you want to use GPT-4 for something, or Grok for another, or Gemini for yet another piece of your workflow? Tough luck. Go open another browser tab, juggle another subscription, copy-paste like an idiot. It's utter nonsense.
I looked at the "bring-your-own-key" (BYOK) solutions out there. And frankly, they're just as bad, but for different reasons. Half of them are so over-engineered. The other half are glorified text boxes that strip away all the productivity features that made Claude's desktop even remotely useful. It's a lose-lose.
I got sick of it. I got sick of the compromises. I got sick of the subscription fatigue. I got sick of feeling like I was being held hostage by some company's API.
So, I did what any sane person would do when faced with a pile of broken tools: I built my own.
This isn't some half-baked wrapper. This is a desktop application built from the ground up to give you the best of both worlds: Claude's "pain-killer" productivity features, but with your keys, your models, and your control.
Here's what I focused on, because these are the things that actually matter for getting work done:
1. A Separate Window for New Threads. Because Context Switching is for Idiots.
You're in Cursor, you're coding. Great. But you need to brainstorm a marketing headline, or research a quick fact, or just have a general chat with an AI without* polluting your code context.
* What do you do now? Open a new browser tab? Switch to a different app? That's inefficient. That's stupid.
My app gives you a *dedicated, separate window** to start a new AI thread. It's always there, always ready. No context switching, no mental overhead. It's a no-brainer.
2. Projects That Actually Work With Your Models. Not Just One.
"Projects" are a good idea. Organize related stuff. Fine. But what if you want to use Claude 3 Opus for the creative writing, then GPT-4 for the technical documentation, and maybe a local Llama for some quick code snippets – all within the same damn project*?
* Current solutions? Forget it. You're stuck.
My solution lets you *seamlessly switch between any model (using your keys) within a single project.** You want to brainstorm with one, refine with another, and generate code with a third? Go for it. It just works. It's about your workflow, not some vendor's limitations.
3. A Plugin System That Isn't a Black Box.
* Everyone's talking about "plugins" and "agents." Most implementations are a mess. You add them, and then what? You have no control.
I built a plugin system that gives you *actual control**:
* Easy Add/Remove: Integrate new tools (MCP servers) without a headache.
* Start/Stop Servers: Don't need that expensive API running all the time? Turn it off. Manage your resources.
* Select Tools: Intuitively choose which tools your AI should use for a specific task. It's not some magic black box; it's a system designed for people who want to control their tools, not be controlled by them.
This isn't some toy. This is a serious desktop environment for serious people who are tired of the current state of AI tools. It's about flexibility, control, and getting actual work done without the constant frustration.
If you're a developer, a writer, a researcher, or anyone who uses AI daily and feels the pain of vendor lock-in and the pathetic state of current BYOK solutions, then you need to check this out.
So, tell me: What's the most infuriating thing about your current AI workflow? And what's one feature you wish existed that actually solves a problem, instead of just being another gimmick? Don't give me platitudes; give me real problems.