Local AI for VS Code and Cursor

Your code. Your models. Your environment.

Reflexor connects VS Code and Cursor to a local AI engine on your computer to understand projects, suggest changes, and run tasks with human review.

Install for Cursor · Install for VS Code

Three reasons to keep AI in your environment.

Project context, local execution, and human approval - without generic marketing cards.

Understands the project

Shows files, symbols, and context retrieved from the open workspace.

Runs in your environment

Local engine, model name, and status stay visible in the workflow.

You stay in control

Diff, command, and approval appear before anything hits the disk.

From editor to local model, with you in control.

No mandatory account and no API key for local models. You confirm downloads and control the environment.

  1. Install Reflexor On VS Code use the Visual Studio Marketplace. On Cursor use Open VSX. VSCodium accepts the same compatible VSIX via Open VSX.
  2. Set up the local engine On first run Reflexor downloads and manages the embedded engine, with confirmation. The binary may download per editor (Cursor vs VS Code).
  3. Download recommended models Models only download with your confirmation: never silent multi-GB pulls. Confirm in chat and start in Ask mode.

Start small. Match your hardware.

Quality and speed depend on the model and your machine. The values below are initial recommendations for this preview.

What changes when AI runs close to your code.

Reflexor prioritizes local inference and control. Cloud assistants prioritize convenience; a manual engine setup offers flexibility with more work.

Reflexor starts as an extension. The vision is a full development experience built around local AI.

Know exactly what to expect before installing.

Privacy, installation, hardware, models, preview, and licensing.

Does Reflexor send my code over the internet?

Local inference uses the AI engine on your computer. Web Search and Web Research are switched off by default. MCP is opt-in. Review the privacy policy before using Reflexor in restricted environments.

Does it work offline?

After the extension, engine, and models are installed, local chat and agent features can work without internet. Engine/model downloads and web features (if enabled) need a connection. We do not claim 100% offline for every scenario.

Which editors and platforms are supported?

Primary editors: VS Code (Marketplace) and Cursor (Open VSX). VSCodium accepts a compatible VSIX, without primary QA. Platforms: Windows x64, Linux x64, and macOS Apple Silicon, no macOS Intel or linux-arm64 in this preview.

Do I need a GPU?

No. The local engine can run models on a CPU, with slower responses. A compatible GPU improves the experience, especially with larger models.

Which model should I try first?

On first run confirm the recommended pack in chat: downloads are never silent. Then compare larger models based on RAM, VRAM, and speed. Models in ~/.ollama are shared between Cursor and VS Code.

Can I review changes before applying them?

Reflexor shows plans, edit proposals, and approval flows. Review every diff and command during the preview.

Is Reflexor free?

Preview 0.3.0 on the release channel is free to use. Future commercial terms will be announced before any change.

Is the source code public?

No. Reflexor is a proprietary product. Open-source dependencies used in the product remain subject to their respective licences and notices.

Demo · FAQ · Privacy Policy · Terms of Use