What is Zed?
Zed is an AI code editor with a built-in AI coding assistant that runs at native speed using Rust and GPU acceleration. It gives developers a fast, low-latency environment where AI for coding works across every layer: completions, agentic editing, and inline transformations. Slow, resource-heavy editors break focus; Zed keeps the entire AI coding assistant experience inside one surface. It runs on macOS, Linux, and Windows, supports external agents via the Agent Client Protocol, and extends through a community extension ecosystem.
Features & Benefits
- Agentic Editing: run an AI coding assistant that reads files, edits code, executes terminal commands, searches the web, and accesses diagnostics autonomously until a task is complete.
- Parallel Agents: run multiple agent threads across several projects simultaneously without switching windows.
- Edit Prediction: get AI code completions on every keystroke using an open-source, open-data model; accept single or multi-line suggestions with Tab.
- Inline Code Transformation: select any code or terminal command, describe the change, and the model rewrites the selection in place with multi-cursor support.
- Multi-Model Support: connect to Anthropic, OpenAI, Google, Ollama, and other LLM providers; run local models, cloud APIs, or both; switch models per task.
- External Agent Support: run Claude Agent, Codex, Gemini CLI, Aider, and other CLI-based agents directly inside the editor via ACP.
- MCP Support: connect MCP servers to extend agent tool access and knowledge.
- Debugger: debug across multiple programming languages using native Debug Adapter Protocol (DAP) support.
- Native Git Support: stage, commit, pull, push, and view diffs directly inside the editor.
- Real-Time Collaboration: co-edit files, share screens, and follow teammates’ cursors using built-in multiplayer infrastructure.
- Remote Development: run the codebase on a remote server while keeping the Zed UI local.
- Language Server Protocol Support: integrate any language with full LSP diagnostics, completions, and hover information.
- Multibuffer Editing: compose excerpts from across the codebase into one editable surface for cross-file work.
- Built-in REPL: execute code interactively through Jupyter kernels without leaving the editor.
- Outline View: navigate a file’s symbols using a structured outline panel.
- Vim and Helix Mode: use full modal editing with bindings, text objects, and marks for both Vim and Helix workflows.
- Extensions: expand language support, themes, and editor functionality through a community-built extension library.
Real-World Applications
A backend engineer working across a large Rust or Python codebase can use Zed’s AI coding assistant to delegate tasks to an agent, watch it navigate files and run commands in real time, then review every change in an editable diff before committing. Parallel agents let them run separate AI for coding tasks across multiple repos without losing focus.
A development team spread across time zones can use Zed’s multiplayer features for synchronous code review. The AI code editor keeps everyone in one surface with shared cursors, co-editing, and screen sharing. The built-in AI coding assistant handles generation and transformation without leaving the editor.
A data scientist or ML researcher prototyping experiments can run code interactively through the built-in REPL and use AI for coding to cut through repetitive boilerplate. The multi-model setup lets them point Zed at whichever model fits the task, or run everything locally through Ollama.
An open-source contributor switching from VS Code or IntelliJ can rely on Zed’s extension ecosystem for language and workflow coverage. Vim and Helix modal editing, LSP diagnostics, and native Git operations give them a complete AI code editor without plugin overhead.
