Void

TL;DR

• What it is: An open-source AI code editor built as a fork of VS Code, offering AI coding features without routing your code through a proprietary backend
• Who it’s for: Privacy-conscious developers, teams using local LLMs, engineers who want direct API control over their AI models
• Cost savings: Cursor Pro costs $20/month per developer; Void itself is free, though you still pay your LLM provider directly — a team of 10 saves $200/month minimum
• Key strength: Direct-to-provider LLM connections — no intermediary server storing your prompts or code
• Key weakness: Development is currently paused; the team has explicitly stated they are exploring “novel coding ideas” and some existing features may stop working

---

What is Void

Void is an open-source AI code editor built as a fork of Microsoft Visual Studio Code. Created by Andrew and Mathew Pareles and backed by Y Combinator, the project positions itself as a direct alternative to Cursor and GitHub Copilot for developers who want AI coding assistance without surrendering their code to a proprietary backend.

The core differentiator is architectural. Cursor and Windsurf route your code through their own servers before sending it to model providers. Void connects directly to whatever LLM you configure — OpenAI, Anthropic, Google, or a locally hosted model via Ollama or LM Studio — without any intermediary storage or logging. “Void sends messages directly to providers without retaining your data,” the README states plainly.

Because it is a VS Code fork, Void inherits the entire VS Code ecosystem: themes, extensions, keybindings, Git integration, and terminal support transfer over in a single click. This removes the primary switching friction that keeps developers locked into VS Code-based tools.

The project launched its first public beta in January 2025 and has accumulated 28,417 GitHub stars under an Apache-2.0 license. However, as of early 2025, the team announced a pause on active IDE development to explore different approaches to AI coding. Features continue working but the project is not under active maintenance.

---

Why people choose it over Cursor, Copilot, and Windsurf

vs. Cursor

Cursor is the category leader for AI code editors and the direct comparison point for Void. Cursor Pro costs $20/month per seat, and all code you write or paste into Cursor passes through Cursor’s servers before reaching the model provider. For teams handling proprietary codebases, regulated data, or security-sensitive projects, this creates a real compliance exposure. Void eliminates that layer entirely.

The tradeoff is maturity. Cursor has a stable release cadence, a large support team, and battle-tested features. Void launched in beta and is currently in maintenance mode.

vs. GitHub Copilot

Copilot is deeply integrated into VS Code and JetBrains IDEs, with organizational controls available in the Enterprise tier ($39/user/month). It does not support local models and routes all context through GitHub’s infrastructure. Void lets you run equivalent functionality against a local Ollama instance at zero recurring cost beyond hardware.

vs. Windsurf

Windsurf (Codeium) offers a competitive free tier but shares the same architectural constraint as Cursor: code travels through their servers. For organizations with air-gapped environments or strict data residency requirements, neither Cursor nor Windsurf is an option. Void, configured with a local model, is.

---

Features: what it actually does

AI coding capabilities

• Tab autocomplete — press Tab to apply inline suggestions
• Quick Edit — inline editing of selected code with natural language instructions
• Chat — context-aware chat with the active codebase
• Agent Mode — can search, create, edit, and delete files and folders, with terminal and MCP tool access
• Gather Mode — read-only variant of Agent Mode that searches but does not modify files

Model flexibility

• Connect directly to any frontier provider: Gemini 2.5, Claude 3.7, GPT-4o, Grok 3, o4-mini, Qwen
• Run fully local models via Ollama or LM Studio: DeepSeek, Llama, Gemma, Qwen
• Custom support for FIM (Fill-in-the-Middle) models used for autocomplete

VS Code compatibility

• Full theme and extension compatibility
• One-click migration of keybindings and settings from VS Code
• Git integration, integrated terminal, all standard VS Code functionality

Additional features

• Checkpoints for LLM-generated changes (review before accepting)
• Lint error detection
• Fast Apply on large files (1000+ lines)
• Prompt transparency — you can inspect the actual prompts being sent to models

---

Pricing math

Option	Monthly cost (10 developers)	Notes
Cursor Pro	$200	$20/seat, proprietary backend
GitHub Copilot Enterprise	$390	$39/seat, VS Code/JetBrains only
Windsurf Pro	$150	$15/seat, proprietary backend
Void + Claude API	~$50-150	Void is free; pay Anthropic directly at usage rates
Void + local Llama	~$0/month	One-time hardware cost for a capable GPU server

The savings are real for teams already paying for API access to model providers. If you are spending $200/month on Cursor Pro and also paying Anthropic for API credits, Void lets you eliminate the Cursor subscription while keeping identical model access.

---

Deployment reality

Void installs like any desktop application. Binaries are available on GitHub for Windows, macOS, and Linux. There is no server to configure, no Docker compose file, no database to initialize.

The complexity appears when configuring model providers. You need to supply API keys directly in the editor settings for cloud providers, or have Ollama running locally with your preferred model already downloaded. For a Llama 3 70B model via Ollama, expect the initial model download to take 20-40 minutes; inference requires a machine with at least 48GB RAM or a capable GPU.

Migrating from VS Code takes roughly 5 minutes using the built-in one-click settings transfer. Most VS Code extensions install without issues. The development pause means some recently introduced VS Code features may lag or break over time.

---

Who should use Void

Best fit

• Security teams or regulated industries (finance, healthcare, defense) where code cannot leave the network
• Individual developers who want AI coding features without a monthly SaaS subscription
• Researchers and tinkerers experimenting with local models
• Teams that already self-host models and want a native integration

Not the right tool if

• You need a stable, actively maintained editor for production use today
• Your team requires onboarding, support contracts, or uptime guarantees
• You want the polished UX of Cursor or the deep enterprise controls of Copilot Enterprise
• You are not comfortable with the possibility that features may break without fixes

---

Alternatives worth considering

• Cursor — The category-defining AI code editor. $20/month, proprietary backend, best overall feature set and polish.
• GitHub Copilot — Deep VS Code and JetBrains integration, $10-39/month. Best if your team is standardized on VS Code and you want Microsoft’s enterprise security controls.
• Continue — Open-source VS Code extension (not a fork) that adds AI features including local model support. More actively maintained than Void, lower disruption to existing VS Code workflow.
• Zed — A new editor built from scratch in Rust with native AI features and real-time collaboration. Much faster than VS Code-based editors but a different extension ecosystem entirely.
• Aider — Terminal-based AI coding assistant that works with any editor. MIT-licensed, actively maintained, works with any Git repository.

---

Bottom line

Void delivers on its core promise: AI coding features with no intermediary server, direct model connections, and full local model support. For privacy-sensitive teams, that architecture is genuinely valuable. The problem is timing — the development team has paused the project to explore new directions, and the current beta is not something you can build a production workflow around. If the team resumes work, Void has a strong foundation. Watch the GitHub repository for signs of life before committing.

Sources

This review synthesizes 5 independent third-party articles along with primary sources from the project itself. Inline references throughout the review map to the numbered list below.

1. [1] dev.to (2024-09-20) — “Cursor AI vs. Void AI Code Editor: The Ultimate Showdown” — comparison (link)
2. [2] itsfoss.com (2025-06-25) — “Void Editor Is Shaping Up Well: Is it Ready to Take on Cursor and Copilot?” — praise (link)
3. [3] infoq.com (2025-06-21) — “The Void IDE, Open-Source Alternative to Cursor, Released in Beta” — news (link)
4. [4] medium.com (2025-04-01) — “Open Source Cursor Alternative Void” — deployment (link)
5. [5] alternativeto.net (2025-04-01) — “Void Editor: Top Alternatives & Analysis” — critical (link)
6. [6] GitHub repository — official source code, README, releases, and issue tracker (https://github.com/voideditor/void)
7. [7] Official website — Void project homepage and docs (https://voideditor.com)

References [1]–[7] above were used to cross-check claims about features, pricing, deployment, and limitations in this review.