Neurix for VS Code
How many tokens? How much will it cost? Does it fit? Will it run on my GPU?
Neurix answers all of it inside VS Code, in 🇺🇸 English and 🇧🇷 Portuguese.
English · Português
✨ Features
- 📊 Live token counter in the status bar: tokens of the selection (or the whole file) and what it costs on your default model, in US$ and R$. Hover for the breakdown and context-window usage, click to compare every model.
- 🧠 Neurix sidebar with all 23 tools, grouped:
- Tokens & cost: count tokens, tokens and cost of a file, cost of one call, compare models, recommend the cheapest model, models and prices, USD → BRL rate.
- Prompt, data & RAG: optimize a prompt (find wasted tokens), cheapest data format (JSON/YAML/CSV/Markdown), chunk a document, RAG vs whole document, tokens per file in a folder, prompt caching savings.
- Budget & alerts: Neurix Budget (monthly cap, projection, the day it runs out, caching and batch) and cost alert (max US$, R$ or tokens).
- Local models: detect GPU/VRAM/RAM, installed Ollama and LM Studio models, does a model fit my GPU, models that fit my GPU, import from Hugging Face, local vs API, real Ollama benchmark.
- Learn: glossary (token, chunk, embedding, quantization, KV cache…).
- 🪄 Context menus: right-click a selection (Editor → Neurix) or a file/folder (Explorer → Neurix). Fields are filled from the selection, the open file or the workspace, and the result appears right away.
- 📋 Results as tables, with Copy JSON, Open JSON and Export Markdown.
- 🤖 MCP server for Copilot: the extension registers the Neurix MCP server, so Copilot agent mode (and other MCP clients in VS Code) can use all 23 tools. Nothing to install: it runs on VS Code's own runtime.
- 🖥️ Interactive CLI: Neurix: Open interactive menu (terminal) opens the animated Neurix menu in the integrated terminal.
Token counts are exact for OpenAI (js-tiktoken) and local models (Hugging Face tokenizers), and estimates for Claude and Gemini (their tokenizers are not public). Neurix always says which.
⌨️ Shortcuts
| Shortcut |
Command |
Ctrl+Alt+T / Cmd+Alt+T |
Count tokens of the selection/file |
Ctrl+Alt+M / Cmd+Alt+M |
Compare models |
Every tool is also in the Command Palette under Neurix:.
⚙️ Settings
| Setting |
Default |
What it does |
neurix.language |
auto |
auto, pt or en |
neurix.defaultModel |
claude-sonnet-5-5 |
Model for the status bar and pre-filled tools (Neurix: Select default model) |
neurix.outputTokens |
300 |
Expected answer size used in estimates |
neurix.iofPct |
0 |
Brazilian IOF tax in % applied to R$ values |
neurix.usdBrl |
live |
Manual USD→BRL rate |
neurix.statusBar.enabled |
true |
Status bar counter |
neurix.statusBar.showCost |
true |
Show cost in the status bar |
neurix.statusBar.maxFileSizeKB |
2048 |
Larger files aren't counted automatically |
neurix.offline |
false |
Don't download price updates or tokenizers |
neurix.mcp.enabled |
true |
Offer the MCP server to Copilot |
Prices update daily from the Neurix repository (24 h cache) and fall back to the table bundled with the extension when offline.
✨ Recursos
- 📊 Contador de tokens na barra de status: tokens da seleção (ou do arquivo inteiro) e quanto custa no seu modelo padrão, em US$ e R$. Passe o mouse para ver os detalhes e o uso da janela de contexto; clique para comparar todos os modelos.
- 🧠 Barra lateral do Neurix com as 23 ferramentas:
- Tokens e custo: contar tokens, tokens e custo de um arquivo, custo de uma chamada, comparar modelos, recomendar o modelo mais barato, modelos e preços, cotação do dólar.
- Prompt, dados e RAG: otimizar prompt, formato mais barato (JSON/YAML/CSV/Markdown), dividir em chunks, RAG × documento inteiro, tokens por arquivo da pasta, economia com prompt caching.
- Orçamento e alertas: Neurix Budget (teto mensal, projeção, dia em que acaba, cache e batch) e alerta de custo (máximo em US$, R$ ou tokens).
- Modelos locais: detectar GPU/VRAM/RAM, modelos instalados no Ollama e LM Studio, cabe na minha GPU?, modelos que cabem na GPU, importar do Hugging Face, local × API, benchmark real com Ollama.
- Aprender: glossário (token, chunk, embedding, quantização, KV cache…).
- 🪄 Menus de contexto: clique com o botão direito numa seleção (Editor → Neurix) ou num arquivo/pasta (Explorer → Neurix). Os campos vêm preenchidos e o resultado aparece na hora.
- 📋 Resultados em tabelas, com Copiar JSON, Abrir JSON e Exportar Markdown.
- 🤖 Servidor MCP para o Copilot: a extensão registra o servidor MCP do Neurix; o modo agente do Copilot usa as 23 ferramentas. Não precisa instalar nada.
- 🖥️ CLI interativo: Neurix: Open interactive menu (terminal) abre o menu animado no terminal integrado.
Configure o idioma com Neurix: Idioma / Language ou neurix.language.
MIT · GitHub · CLI: npm install -g aspael-neurix
| |