Zed Review

Zed is a native editor built for speed using multiple cores and the GPU. Features include multibuffer editing across excerpts of your codebase, native Git support (stage, commit, diff, pull, push), a debugger built on the Debug Adapter Protocol, full LSP support, Vim and Helix modes, a built-in REPL through Jupyter kernels, and Dev Containers. The AI docs describe an agent panel where you prompt agents, add context, review changes, and manage threads, with Parallel Agents running multiple threads across projects and worktrees, an inline assistant that transforms a selection in place, edit prediction (Zeta) that offers completions as you type, external agents through the Agent Client Protocol that run in their own process and configuration, MCP servers, and built-in tools with profiles, skills, and instructions. The Zed Agent can use Zed-hosted models, provider API keys, supported subscriptions, gateways, and local models, and separate pages cover AI privacy (provider data boundaries) and opt-in sharing of feedback and training data. Extensions can add languages, themes and icon themes, snippets, MCP servers, and debug adapters.

Zed is a desktop application you download from zed.dev or install through a local package manager on macOS, Linux, or Windows (on Windows, winget install -e --id ZedIndustries.Zed, or the preview build). Windows requires a DirectX 11 compatible GPU, and in a virtual machine it uses the emulated adapter and may be slower. Remote development splits the work: the UI, language model calls, Tree-sitter parsing, and unsaved changes run locally, while source code, language servers, tasks, and terminals run on the remote server. Zed uses your local ssh binary and a ControlMaster connection, checks ~/.zed_server on the remote for a server binary that exactly matches your local Zed version (downloading it on the remote, or uploading it when upload_binary_over_ssh is enabled), and reattaches to a server daemon so a dropped connection can resume, with unsaved changes persisted locally. Requirements are Zed v0.159 or later, SSH access, and a remote machine on macOS Catalina or later, Linux x86_64 or arm64 (32-bit not supported), or Windows x86_64 or arm64, plus WSL folders. The README links platform-specific build guides, and the source is GPL-3.0-or-later with Apache-2.0 components where marked.

The editor is minimal and keyboard-driven, with panels for the agent, Git, and the debugger, and a multibuffer view for editing excerpts from many files in one place. Collaboration lets you share a project and work in the same buffers. Remote development has a few documented rough edges: you cannot open files from the remote terminal with the zed command, very large directories (for example / or ~ with more than 100,000 files) are not handled well so you should open specific projects, and the old relay mode through Zed's servers was removed in v0.157. We did not use Zed hands-on; this section reflects the docs, repository, pricing page, and OpenAlternative.

• GPU-accelerated, multi-core native editor written in Rust
• Parallel AI agents across projects and worktrees, an inline assistant, and Zeta edit prediction
• External agents through the Agent Client Protocol, plus MCP servers
• Multiplayer editing and screen sharing
• SSH and WSL remote development with a version-matched server binary
• Debugger on the Debug Adapter Protocol, full LSP, Git, and a Jupyter-based REPL
• Vim and Helix modes, Dev Containers, and extensions for languages, themes, snippets, and debug adapters

• Speed and responsiveness that Electron-based editors struggle to match
• Flexible AI: use hosted tokens, your own keys, Ollama, subscriptions, or external agents
• Collaboration is a core feature rather than a plugin
• Active development, with about 92,000 stars and commits within the hour of this review

• The free plan includes 2,000 accepted edit predictions; unlimited predictions need Pro
• Pro includes $5 of hosted AI tokens, and usage beyond that is billed at API list price plus 10%
• Business SSO, SAML, and SCIM are described as planned but not yet available
• Windows needs a DirectX 11 GPU, and remote development requires an exact local and remote version match
• The extension ecosystem is smaller than VS Code's, and the extension docs list capabilities (languages, themes, snippets, MCP servers, debug adapters) rather than UI extensions
• GPL-3.0-or-later licensing for the editor, which matters if you want to embed it

Personal is free and includes 2,000 accepted edit predictions, and you can use Zed with your own API keys (Amazon Bedrock, Anthropic, GitHub Copilot, Google AI, Ollama, OpenAI, OpenRouter, and others) or external agents such as Claude Agent and Codex CLI. Pro is $10/month with unlimited edit predictions and $5 of hosted AI tokens, with extra usage billed at API list price plus 10% and an optional spending cap, plus a 14-day trial with $5 of usage and no credit card. Business is $30/seat/month with org-wide AI model policies, data governance controls, spend visibility, and role-based access, with order forms at 25 or more seats, no free trial, and no bundled AI credits per seat.

Cursor is a proprietary AI code editor forked from VS Code, while Kiro and Antigravity are other AI-first IDEs. Zed is open source, native, and much lighter, with AI that you can power with your own keys or external agents, and collaboration built in. Cursor has a larger extension ecosystem through its VS Code base, and Zed trades that for speed and openness. If you already rely on specific VS Code extensions, check that Zed covers them before switching.

• Developers who want a fast, native editor with built-in collaboration
• Teams that want to use their own model keys or external agents inside the editor
• Rust, Go, and TypeScript developers who value low latency and remote development over SSH

• Developers who depend on specific VS Code extensions
• Teams that need SSO and SCIM today
• Anyone who wants a web-based IDE