NeuroSwitcher
New in the Pro beta · AI text assistant
NeuroSwitcher brings a language model to every text field in Windows — email, chat, browser, IDE. Rewrite a clumsy sentence in place, translate a paragraph, summarize a wall of text, or ask a quick question without switching windows. Run it on free NeuroSwitcher Cloud, your own API key, or a local model — your choice.
hey so um we cant make the deadline sorry, next week ok??
Unfortunately we won't make the deadline. Would next week work for you?
Two hotkeys
No new app to live in, no browser tab with a chat. The assistant meets your text where it already is.
Select text, press the hotkey, pick a preset — Tidy, Formal, Casual, a translation, or your own instruction — and the result is typed right over the selection. Fix tone, grammar, or language without leaving the app you're in.
Ask a free-form question — about the selected text or about anything. The answer opens in a small mini-chat window with a follow-up box and Copy / Replace buttons. Close it, and the conversation is gone; nothing accumulates. You can also attach screenshots and images to the question — not just text.
Both hotkeys are configurable, and the whole feature can be toggled from the tray. While a request runs, a small "⟳ LLM…" badge shows next to your cursor.
Screenshots & images
The Ask window takes pictures as well as text. Attach a screenshot of an error dialog, a chart, a photographed document, or a UI you're stuck on — and ask about it in the same little window.
Hit the scissors button and drag a box around any part of the screen. The window steps out of the way while you capture, then comes back with your shot attached.
One hotkey grabs a region and drops you straight into the Ask box with the screenshot ready — perfect for "what does this error mean?" without a single extra click.
Pick one or more images (PNG, JPG, WebP, BMP, GIF) from disk. Text files (.txt, .md) come in as extra context for your question.
Copy an image anywhere, then press Ctrl+Shift+E — it's staged as a removable attachment automatically, so a quick snip tool flows right in.
A "🖼 n" counter shows what's attached; click it to clear. Nothing is sent until you press Send. Images need a vision-capable model — NeuroSwitcher Cloud's works out of the box, as do vision models on Ollama (llava, qwen2.5-vl…), OpenAI, or OpenRouter. Attach one image per question on the free tier, up to five with a Pro license.
Presets
The launcher offers ready-made presets, and every prompt behind them is editable in Pro Console — you can also add your own.
Same message, different register. Turn a hasty note into a professional reply — or loosen up stiff prose.
English, Russian, German, Hebrew — matching the languages NeuroSwitcher already lives in. Select, hotkey, translated.
Condense a long thread to its key points, or have a dense paragraph explained in plain words.
Reshape rambling text into a clean list of points — handy for meeting notes and tickets.
Select a question someone sent you, and get a drafted answer to edit rather than a blank page.
A free text field for anything else: "make this half as long", "turn this into a polite refusal", "explain this regex".
Bring your own model
The assistant speaks the standard OpenAI-compatible chat API, which means you decide where the intelligence lives.
The zero-setup default: a free daily quota with no account, no key and nothing to install. Your text is relayed through NeuroSwitcher's proxy so you can start immediately — switch to a local model any time you want nothing to leave your PC. More about NeuroSwitcher Cloud →
Run open models like Qwen or Llama locally with Ollama. NeuroSwitcher finds it on localhost — no key, no account, no internet needed.
Prefer a GUI for managing local models? LM Studio's local server works out of the box the same way.
Want frontier-model quality? Paste your own API key. Requests go straight from your PC to that provider — on this path there is no NeuroSwitcher relay, and we never see your key or your text.
A custom URL setting accepts anything that speaks /chat/completions — a home server, a company gateway, a niche provider. Per-provider models and keys are remembered separately.
Every backend is opt-in — the assistant does nothing until you enable it and pick one. NeuroSwitcher Cloud uses an anonymous free daily quota; your own keys are stored locally on your machine; a local model keeps everything on your PC.
Better together
NeuroSwitcher's local voice input can pass the raw transcript through the same LLM before typing it — punctuation fixed, filler words gone, and optionally reshaped by any preset. Dictate a stream of thought, watch a readable paragraph come out.
FAQ
No. Start on free NeuroSwitcher Cloud (a daily free quota, nothing to set up) or free local models via Ollama or LM Studio — no account, no key, no cost. For frontier quality, bring your own OpenAI or OpenRouter key and pay the provider directly at API rates.
Yes. With a local backend the entire flow — selection, rewrite, mini-chat — runs on your machine. This is the same local-first principle as the rest of NeuroSwitcher.
Only what you explicitly submit: the selected text and/or your typed request, any images you attached, plus the preset's instruction. It travels directly from your PC to the provider you configured, under your key. NeuroSwitcher's own servers are never involved, and your regular typing is never sent — the assistant only acts when you press its hotkey.
Yes. In the Ask window you can snip a screen region (✂), attach image or text files (📎), or press Ctrl+Alt+A to grab a region and ask about it in one step; a copied image is picked up automatically when you press Ctrl+Shift+E. Images need a vision-capable model — NeuroSwitcher Cloud works out of the box, as do vision models on Ollama, OpenAI or OpenRouter. Attach one image per question on the free tier, up to five with a Pro license, and nothing is sent until you press Send.
For rewriting and cleanup, small instruct models (1–8B, e.g. Qwen 2.5 or Llama 3.x class) are fast on ordinary hardware and do a solid job. For heavier tasks — long summaries, tricky translations — a larger local model or a cloud model gives better results.
It works wherever text can be selected and typed: browsers, mail clients, chats, editors, IDEs, office apps. Rewrite-in-place needs a selection; Ask works even with nothing selected.
It ships as part of the free Pro beta. NeuroSwitcher Cloud has a free daily quota, and a local model via Ollama or LM Studio costs nothing at all. If you connect your own cloud key, you pay only your provider's API prices.
Free Pro beta
Download NeuroSwitcher for Windows 10 and 11, use free NeuroSwitcher Cloud, your own API key, or a local model via Ollama, and rewrite, translate, and ask — anywhere you type.
Download NeuroSwitcher