Void — Open-Source Cursor Alternative

What it is

Void is an open-source AI code editor that positions itself as an alternative to Cursor. Built on VS Code, it provides AI-assisted coding features including inline completions, chat, and multi-file editing. Void differentiates from Cursor by being fully open source (Apache 2.0), supporting custom and local models, and providing a checkpoint system for saving and restoring editor states.

Void is designed for developers who want Cursor-like AI coding features but prefer open-source tools or need to use custom/local LLM models.

How it saves time or tokens

Void brings AI coding features into a familiar VS Code interface without vendor lock-in. You choose your LLM provider (OpenAI, Anthropic, local models via Ollama) and keep full control over where your code is sent. The checkpoint system lets you save editor state before large AI-driven changes, making it safe to experiment with aggressive refactoring. If the AI makes a mistake, revert to a checkpoint instead of manually undoing changes.

How to use

1. Download Void from voideditor.com or build from source:

git clone https://github.com/voideditor/void.git
cd void
# Follow build instructions in README

1. Configure your LLM provider in Void settings (OpenAI, Anthropic, Ollama, or custom endpoint).

1. Use AI features:

Cmd+K       — Inline AI edit
Cmd+L       — Open AI chat
Cmd+Shift+K — Create checkpoint
Cmd+Shift+Z — Restore checkpoint

Example

Using Void with a local Ollama model for private coding:

# Start Ollama with a coding model
ollama run codellama:34b

# In Void settings:
# Provider: Ollama
# Model: codellama:34b
# Endpoint: http://localhost:11434

# Now all AI assistance runs locally
# No code leaves your machine

This setup provides AI coding assistance with complete privacy, as all inference happens on your local hardware.

Related on TokRepo

• Coding tools — Browse AI code editors and assistants
• Local LLM tools — Explore local model inference options

Common pitfalls

• Expecting feature parity with Cursor on day one. Void is newer and actively developing. Some features available in Cursor may not yet be implemented in Void.
• Not configuring an LLM provider before trying AI features. Void does not include a bundled AI model. You must configure at least one provider (cloud or local) in settings.
• Using a local model that is too small for coding tasks. Models under 13B parameters generally produce poor code suggestions. Use at least a 34B parameter model (like codellama:34b) for acceptable quality.