Local LLM Plugin — AI Assistant for Visual Studio
A fast, private, beautifully designed AI coding assistant that lives inside Visual Studio.
Run models locally with Ollama or LM Studio, or use OpenAI, Anthropic and Google Gemini — with streaming answers, one-click quick actions and a single-step undoable apply, all in a macOS-inspired Liquid Glass interface.
Version 4.0 is a complete refresh: new interface, live streaming, conversations, quick actions, secure key storage, Visual Studio 2026 and Arm64 support — and many fixes. See the release notes.
Why you'll like it
- Private by default — use local models through Ollama, LM Studio, llama.cpp, vLLM, Jan or LocalAI and nothing leaves your machine.
- Your choice of cloud — OpenAI, Anthropic and Google Gemini are one setting away. Every provider keeps its own URL, model and API key.
- Answers as they are written — responses stream token by token; press Esc to stop at any time.
- Built for code — only the code is applied to the editor, indentation is matched, line endings are kept, and the whole change is a single Ctrl+Z.
- Looks at home on any theme — dark and light Liquid Glass themes follow your Visual Studio theme, with real Windows 11 acrylic blur.
Features
Ask, refine, apply
- Streaming chat with follow-up questions — the conversation keeps its context. New chat with Ctrl+N.
- Quick actions in one click: Explain, Refactor, Fix bugs, Optimize, Document, Unit tests, Review. Type extra instructions first to steer them.
- Code context — the selection, file name and language are attached automatically. Edit the code before sending, or turn on Whole file to send the entire document.
- Markdown answers with headings, lists and inline code, plus code cards with their own Copy and Apply buttons.
- Thought process — reasoning from thinking models (DeepSeek-R1, Qwen3, gpt-oss…) is shown separately instead of mixed into the answer.
- Regenerate an answer, or Use as input to keep transforming a result.
Apply exactly what you want
| Action |
Shortcut |
What it does |
| Apply |
Shift+Enter |
Replaces the selection with the code from the latest answer (explanations and Markdown fences are stripped). |
| Insert below |
Ctrl+Shift+Enter |
Inserts the result below the selection — ideal for tests and docs. |
| Partial apply |
highlight + Shift+Enter |
Applies only the part of an answer you highlighted. |
| Per block |
Apply on a code card |
Applies one specific code block. |
Every edit uses the Visual Studio editor API as one undo transaction, re-indents the code to fit where it lands and uses the document's line endings.
Text tools — no AI required
Remove duplicate lines · Modify & replicate · Erase text · Add to every line · Sort A→Z / Z→A · Reverse lines · Remove empty lines · Trim trailing whitespace · UPPERCASE · lowercase.
Results keep your line endings and can be applied, copied or chained into the next tool.
Supported providers
| Provider |
Runs |
API key |
| Ollama |
Locally |
– |
| LM Studio / any OpenAI-compatible server (llama.cpp, vLLM, Jan, LocalAI…) |
Locally |
Optional |
| OpenAI |
Cloud |
Required |
| Anthropic |
Cloud |
Required |
| Google Gemini |
Cloud |
Required |
Press ↻ in Settings to list the models installed in Ollama or available to your API key, and Test connection to check everything in one click.
Getting started
- Install the extension and restart Visual Studio.
- Local: install Ollama and run
ollama pull qwen2.5-coder:7b.
Cloud: open Settings (Ctrl+,), choose a provider and paste your API key.
- Select some code and press Ctrl+1 — or right-click in the editor and choose AI Assistant (Local LLM), or use Tools ▸ AI Assistant (Local LLM).
- Pick a quick action or type a question, then press Shift+Enter to apply the result.
Keyboard shortcuts
| Shortcut |
Action |
Ctrl+1 |
Open the assistant |
Enter / Ctrl+Enter |
Send / new line |
Shift+Enter |
Apply (replace selection) |
Ctrl+Shift+Enter |
Insert below |
Esc |
Stop streaming · close |
Ctrl+N / Ctrl+, |
New chat / Settings |
Privacy & security
- With local providers, your code never leaves your computer.
- API keys are encrypted with Windows DPAPI for your user account and stored per provider in
%AppData%\LLLMPluginModern\settings.json.
- Requests go only to the endpoint you configure. Local addresses bypass the system proxy; cloud requests honor it.
Compatibility
Visual Studio 2022 (17.0 and later) and Visual Studio 2026 — Community, Professional and Enterprise — on x64 and Arm64.
Settings from version 3.x are migrated automatically.
Feedback
Questions, ideas and bug reports are welcome on GitHub.
If the extension helps you, a ⭐ rating on the Marketplace is much appreciated!
🇹🇷 Türkçe
Visual Studio için hızlı, gizliliğe saygılı ve şık bir yapay zekâ kodlama asistanı. Modelleri Ollama veya LM Studio ile yerelde çalıştırın ya da OpenAI, Anthropic ve Google Gemini kullanın. Canlı akış hâlinde yanıtlar, sohbet, tek tıkla hızlı eylemler (Açıkla, Refactor, Hata düzelt, Optimize et, Belgele, Birim testi, İnceleme), editöre tek Ctrl+Z ile geri alınabilen uygulama, 10'dan fazla metin aracı ve Visual Studio temasını takip eden macOS tarzı Liquid Glass arayüz. API anahtarları DPAPI ile şifrelenir. Visual Studio 2022 ve 2026, x64 ve Arm64 desteklenir. Ayrıntılar: GitHub.
| |
