Zed

Zed is a code editor built from scratch in Rust by Nathan Sobo, Antonio Scandurra, and Max Brunsfeld — the team behind Atom, Electron, and Tree-sitter. It is available for macOS, Linux, and Windows, and its core source code is publicly hosted on GitHub. The editor emphasizes raw performance through GPU-accelerated rendering and multi-core CPU utilization, while also offering native multiplayer collaboration and deep AI integration.

What It Is

Zed is a desktop code editor in the same category as VS Code, Neovim, and JetBrains IDEs, but differentiated by its Rust-native architecture and first-class support for real-time collaboration. The homepage describes it as "a minimal code editor crafted for speed and collaboration with humans and AI." Unlike editors that bolt on AI features via extensions, Zed integrates agentic editing, inline AI assistance, and edit prediction directly into the core product. It supports Language Server Protocol (LSP) for language intelligence, Tree-sitter for syntax parsing, and the Debug Adapter Protocol (DAP) for native debugging.

Architecture and Performance

Zed is written entirely in Rust and uses a custom UI framework called GPUI that renders every frame on the GPU. The team states the editor is designed to "efficiently leverage multiple CPU cores and your GPU." This architecture is the primary reason the editor is positioned as faster than Electron-based alternatives. The project has accumulated over 83,000 GitHub stars and more than 8,500 forks as of mid-2026, with 1,623 contributors and 841 pull requests merged in a single recent month according to the homepage.

AI and Agentic Editing

Zed's AI layer is built into the editor rather than delivered as a plugin. Key AI capabilities include:

• Parallel Agents: Run multiple AI agent threads across several projects simultaneously without leaving the editor flow.
• Agentic Editing: Delegate file edits, code navigation, and tool use to an AI agent and review changes live.
• Edit Prediction: Powered by Zeta2, described as an open-weight language model that predicts the developer's next edit.
• Inline Assistant: Send selected code to a language model for in-place transformation.
• Any Agent, Any Tool: Supports external agents like Claude Agent, Codex, and OpenCode via the Agent Communication Protocol (ACP), and MCP servers for extending agent knowledge.

Supported AI providers include Anthropic, OpenAI, Google AI, Amazon Bedrock, GitHub Copilot, Deepseek, Mistral, Ollama, OpenRouter, Vercel, LM Studio, and others. Users can bring their own API keys or use Zed-hosted models.

Collaboration and Multiplayer

Real-time multiplayer is a core feature, not an add-on. Developers can share projects, follow collaborators' cursors across files, co-edit in the same buffer, and share their screen — all without leaving the editor. The collaboration model is described as aiming to "eliminate the distinction between local and remote projects as much as possible." Guests can open, edit, save files, search, and interact with the language server in a shared session.

Open-Source Model and Extensions

Zed's source code is publicly available on GitHub under a mixed license (the repository lists "Other/NOASSERTION," which covers multiple licenses for different components). The homepage states it is "built by a global, growing, and thriving community of thousands of developers." An extensions ecosystem provides language support, themes, and tooling — popular extensions include HTML (5.4M installs), TOML (1.1M), Catppuccin themes (884k), and Java (866k), among hundreds of others. Developers can create and publish their own extensions.

Update: v1.2.7

The latest stable release is v1.2.7, published on May 19, 2026. Recent blog posts from May 2026 cover local model support, ChatGPT subscription integration, and changes to Anthropic's Claude billing as it affects Zed users. The homepage prominently features Parallel Agents as the newest major capability, with a dedicated blog post and landing page. The roadmap and weekly release cadence signal active, ongoing development across performance, AI, debugging, and remote development features.