Goodbye API Keys and Paid Subscriptions — Proxima Combines ChatGPT, Claude, and Gemini in One Window
Recently stumbled upon Proxima — a project that solves one of the most annoying problems in modern development. Familiar situation: you have a ChatGPT Plus subscription, you're used to Claude for writing code, and you use Perplexity for quick web searches. But the moment you need to integrate into an IDE or automate through scripts, the circus starts. Either you pay per token through the API (which adds up fast), or you constantly switch between browser tabs.
Proxima offers a damn elegant workaround that turns your regular accounts into a full-fledged local API gateway.
What Is This Thing Anyway
In short, Proxima is a local Electron-based server that "pretends" to be a browser. It uses your active sessions in ChatGPT, Claude, Gemini, and Perplexity. You log in to the app once, and then Proxima bridges your code editor and these chats.
No API keys, no token limits (except those in your subscription), and no per-request payments. All traffic goes through 127.0.0.1, so your data doesn't fly off to some intermediary — it goes directly to the AI model you trust.
How It Works Under the Hood
In the new version 4.1.0, the developers rolled out the Provider Engine system. Previously, similar tools often relied on "home parsing" (DOM scraping), which was slow and broke with any website interface update.
Proxima took a different route. It injects scripts directly into the browser page context and communicates with provider APIs at the SSE stream level. This is an order of magnitude more stable and faster — the response is almost instantaneous, as if you were using the official API, but through your own account. If the engine can't hook into the page's internal methods for some reason, it falls back to the good old method of interacting with interface elements.
What Proxima Can Do
The project developer Zen4-bit clearly understood that a simple "chat in a window" nobody needs, so they loaded it up with features to make it a proper Swiss Army knife.
1. Unified OpenAI-compatible endpoint
This is probably the main feature. The app spins up a local server on http://localhost:3210. Any tool that can work with the OpenAI API (for example, extensions in VS Code or Cursor) can be redirected to this address. Just change the URL in the plugin settings, and voila — your code editor now talks to Claude 3.5 Sonnet through your regular account.
2. Smart router mode
You can skip choosing a specific model and send a request to model: "auto". The system will decide on its own that a coding task is better suited for Claude, and searching for the latest news should go to Perplexity. There's even a model: "all" mode, where the same question goes to all available AI models at once, and you sit there comparing results in one window.
3. Powerful CLI
For those who don't like leaving the terminal, there's the proxima utility. It lets you pipe compilation errors or test output directly to the AI:
npm run build 2>&1 | proxima fix
Or quickly ask for code advice:
proxima code review --file src/auth.js
4. MCP support
For Model Context Protocol (MCP) fans, there's plenty to explore here. The project provides over 45 tools: from deep GitHub and Reddit search to automatic test writing and security auditing.
How to Get It Running
On Windows, it's straightforward — download the installer from releases and run it. For macOS and Linux, you'll need to get your hands dirty with the terminal:
- Clone the repository:
git clone https://github.com/Zen4-bit/Proxima.git - Install dependencies:
npm install - Launch:
npm start
After launching, an Electron window will open. There you need to log into your ChatGPT, Claude, or Gemini accounts. Note that Google sometimes blocks login through embedded browsers, so for Gemini you often have to use cookies from your main browser — the app includes instructions on how to do this.
Practical Use Cases
Why bother if you can just open a tab in Chrome?
First, automation. You can write a Python script that goes through your local files and asks Claude to update the documentation, without spending money on the Anthropic API. Proxima comes with ready-to-use Python and JavaScript SDKs.
Second, context. Thanks to MCP tools, the AI gets access to things that are hard to just copy-paste into a chat: project structure, internet search results, or specialized code analysis utilities.
Third, savings. If you code a lot with AI assistants, the API bills at the end of the month can be an unpleasant surprise. Proxima lets you use your personal subscription resource at 100%.
Is It Worth Trying
The project looks very much alive and is actively updated. Of course, there's always the risk that OpenAI or Anthropic will tighten the screws and start fighting these access methods more aggressively. But right now, it's one of the most convenient ways to gather all the top AI models in one local port.
Especially recommend Proxima for those who work in Cursor or VS Code and want to use different models without juggling subscriptions and API keys. The project is free for non-commercial use, so dropping a star on GitHub is definitely not amiss.
The only downside is that the documentation is somewhat sparse in places, and you have to poke around in the app settings to figure out which port goes where. But the interactive docs at http://localhost:3210/docs after launching the server close that gap.
Related projects