I stopped paying for multiple AI coding tools and built one setup that can use Claude, GPT, Gemini, and local models

Published Sep 27, 2026, 6:00 PM EDT Mahnoor Faisal is a tech journalist covering AI and productivity tools with bylines at XDA, SlashGear, MakeUseOf, Laptop Mag, and Android Police. She's been writing professionally since she was sixteen, and has since penned hundreds of articles. This includes in-depth coverage of AI tools like NotebookLM to breaking news across the AI space. Her passion for technology started when she received her first iPod Touch (4th generation) on her 8th birthday, and she's been deep in the tech world ever since. Currently pursuing a degree in computer science, Mahnoor brings both a journalist's eye and a technical foundation to her coverage of how AI is reshaping the way we work and learn. One thing I've realized, beyond the fact that AI tools are only getting more expensive by the day, is that every LLM has different strengths. One might be excellent at frontend work, while another is far better at debugging, reasoning through a messy codebase, or handling a completely different kind of task. Beyond code, a model might be great at reasoning but fall short when it comes to writing, research, or working with large amounts of context. The point here is, you can't really crown one model the best at everything. The one I want to use depends almost entirely on what I'm trying to get done. However, paying for each subscription just so I can bounce between the models I prefer gets expensive very quickly. This is why I stopped paying for multiple AI coding tools and built one setup that lets me use Claude, GPT, Gemini, and even local models from the same place. The problem with AI subscriptions is that you’re still locked in One subscription only gets you so far While I've already touched on this above, I wanted to spend a little more time elaborating on what I mean by this. The problem isn't necessarily that any one AI subscription is bad value. It's that subscribing to one usually means committing yourself to that company's models if you want to get the most out of what you're paying for. If I'm already paying for Claude, for example, it makes sense to use Claude Code. The same logic applies to ChatGPT and OpenAI's coding tools, or Gemini and Google's own ecosystem. But the moment I want to use a different model because it's better suited to a particular task, I'm suddenly looking at another subscription. When it comes to local LLMs, though I don't need to pay for a separate subscription by any means, I still usually need to switch to another app or workflow entirely. That ends up creating the same problem in a different form: instead of juggling subscriptions, I'm juggling interfaces. OpenCode gave me the best of both worlds Why choose one model when I can have them all? What I really wanted was one place where I could pick whichever model made the most sense for the task right in front of me. Given that most of what I use these models for is coding anyway, I figured a coding tool was likely the way to go. That also made more sense to me than starting with a general-purpose AI app. Coding tools can still handle plenty of non-coding tasks with ease, whether that's explaining something, brainstorming an idea, or working through a problem. The reverse isn't quite as true. A general chatbot might be able to write code, but that doesn't suddenly give it the project awareness, file access, terminal integration, and agentic workflow of a dedicated coding tool. This is where OpenCode comes in. It's an open-source AI coding agent that sits in roughly the same category as tools like Claude Code and Codex. Beyond having source code that anyone can inspect, though, the biggest reason it caught my attention is that it doesn't tie itself to one company's models. OpenCode supports more than 75 LLM providers, including the major names I'd actually want to switch between, and it can connect to local models as well. So instead of choosing a coding tool based on whether I want Claude, GPT, or Gemini that day, I can keep the interface the same and simply swap the model underneath it. My setup mixes APIs, local models, and one subscription My wallet likes this setup a lot more I didn't go fully subscription-free, and I don't think that's necessary for this setup to make sense. Instead, I ended up with a mix of one subscription, a few API keys, and local models depending on what I actually want to use. For the models I only reach for occasionally, API access makes far more sense than paying for another monthly plan. I can add the provider to OpenCode, use the model when I need it, and only pay for the usage I generate. That works particularly well for models I like having access to but don't use enough to justify keeping another subscription active every month. While I was initially torn between keeping Claude Code and ChatGPT Plus, OpenCode made that decision surprisingly easy. It supports signing in with a ChatGPT Plus or Pro subscription directly, which means I can use the OpenAI models I'm already paying for inside OpenCode itself. Claude doesn't work quite the same way. Anthropic doesn't officially allow Claude Pro or Max subscriptions to be used with OpenCode, so if I want Claude in the same setup, I need to use Anthropic's API instead. That actually works out pretty well for me. I can keep ChatGPT Plus as the one subscription I'm paying for, use Claude through the API when I specifically want it, and do the same with Gemini or any other provider I only reach for occasionally. Then, for lighter tasks, I can fall back to local models without paying anything at all. The end result is that I'm not really giving anything up. I still have access to the models I actually want to use, but I'm no longer maintaining a separate monthly subscription for every coding tool just to get there. OpenCode simply becomes the layer that ties all of those options together. The model flexibility is only part of why OpenCode works The ability to switch between models is what initially drew me to OpenCode, but it helps that the rest of the experience feels like a proper coding agent too. It can work across a project, edit files, run terminal commands, and use LSP integration to better understand the codebase instead of treating everything like plain text. I also like that it gives me some control over how autonomous it gets. OpenCode has separate Plan and Build modes, along with granular permissions for things like file edits and shell commands. It also supports MCP servers if I want to extend it further. So while model flexibility is the main reason I use it, I don't feel like I'm giving up the features I'd expect from a dedicated coding tool just to get that freedom.