Getting Started With Zed Editor and Its AI Features
Zed is a high-performance code editor built in Rust, designed from the ground up for speed and real-time collaboration. Unlike most editors that bolt on AI features as plugins, Zed ships with both built-in multiplayer editing and an AI assistant called Zed AI – making it a genuinely different working environment for teams who want those two things to actually work together.
Installing Zed and Configuring the Basics
Zed is currently available for macOS and Linux. Head to zed.dev and download the appropriate installer for your system. On macOS, it installs like any standard app – drag to Applications and launch. On Linux, Zed provides an install script you run directly in the terminal: curl https://zed.dev/install.sh | sh. The binary installs to ~/.local/bin/zed by default, so make sure that path is included in your shell’s PATH variable.
Once Zed is open, the first thing to address is your settings file. Press Cmd + , (macOS) or Ctrl + , (Linux) to open settings.json. This is where almost everything lives – theme, font size, keybindings, language server preferences, and AI configuration. Zed uses a flat JSON structure here, so it is straightforward to read and edit directly. If you are migrating from VS Code, the mental model is similar but without extension marketplace sprawl.
Set your preferred font and theme early. Zed supports Zed Themes natively, and a growing number of community themes are available through the extensions panel, accessible via Cmd + Shift + X. For font rendering, the editor performs best with a monospace font that has good ligature support – something like JetBrains Mono or Fira Code works well. Add "buffer_font_family": "JetBrains Mono" and "buffer_font_size": 14 directly to your settings file to apply them.
Language servers are enabled automatically for most common languages when you open a relevant file. Zed uses the Language Server Protocol, so support for TypeScript, Rust, Python, Go, and others is handled without extra setup in most cases. For Python specifically, you may want to point Zed at a virtual environment by adding "python": { "default_language_server": "pyright" } inside the "languages" key in settings. Pyright gives you strong type inference and import resolution, which matters when working in larger codebases with multiple contributors.
Setting Up Zed AI and Using the Assistant Panel
Zed AI is the editor’s built-in AI assistant, powered by Anthropic’s Claude models. To activate it, you need a Zed account – sign up at zed.dev and then sign in from within the editor using Cmd + Shift + P and searching for “sign in.” Once authenticated, Zed AI activates automatically and you can open the AI assistant panel with Cmd + ? on macOS or by clicking the assistant icon in the right toolbar.
The assistant panel operates as a persistent conversation thread alongside your code. You can type questions, paste error messages, or ask it to explain a function directly from context. What makes this more useful than a browser-based chat is the inline context injection – highlight a block of code, press Cmd + Enter, and Zed sends that selection along with your prompt to the model. The response appears in the panel, and you can accept edits directly into the file with a single click. There is no copy-pasting back and forth.
For inline code suggestions, Zed supports a feature called Edit Mode. With your cursor inside a function or block, trigger it via Cmd + Shift + Enter and describe what you want the code to do differently. The model returns a diff-style suggestion that Zed overlays directly in the editor buffer. You see what changes, what gets removed, and what stays – then accept or reject the whole patch. This is meaningfully different from autocomplete; it is more like asking a colleague to take a pass at a specific section while you watch.
Context management matters when working on large projects. Zed AI lets you pin specific files or symbols to a conversation thread so the model always has them in scope. Open the assistant panel, click the + Add Context button, and select files from your project tree. You can also drag a file tab directly onto the assistant panel to add it. Pinning your main types file, a configuration module, or a shared utility library means the AI’s suggestions will stay consistent with your actual codebase rather than making assumptions about how things are structured.
If you prefer to use a different AI provider – OpenAI, for example, or a locally running model through Ollama – Zed supports custom provider configuration. In your settings.json, add an "assistant" block with the provider type and your API key or endpoint URL. A local Ollama setup pointing to http://localhost:11434 works well for teams with data residency requirements or anyone who wants AI assistance without sending code off-device. The tradeoff is that smaller local models are noticeably less capable on complex refactors, so the choice depends on what your team is optimizing for.
Using Real-Time Collaboration With a Team
Zed’s multiplayer feature is built directly into the editor rather than layered on top. To start a session, open a project and press Cmd + Shift + P, then run “Share Project.” Zed generates a link that invited collaborators can follow – they join your session inside their own local Zed instance, seeing your full project tree and editing in real time. Cursors are labeled with each participant’s display name, and the experience is closer to Google Docs than to anything a traditional IDE has offered. You can see exactly where someone else is looking, not just what they last committed.
Permissions within a session are worth configuring before you invite others. The host controls who can edit versus view, and you can grant or revoke write access per user from the collaboration panel. For code review sessions, read-only access for one participant while another walks through changes is a natural fit. For pair programming, both users get full edit rights and either can drive at any time. One practical note: if your team is working behind a strict firewall or corporate VPN, Zed’s peer-to-peer connection may need an exception, since it does not route through a central relay server by default.
