Key Takeaways
- Built in Rust by the creators of Atom) and Tree-sitter, using Zed's own GPUI framework for GPU-accelerated rendering
- License is not a single term: GPL-3.0 for the core editor, AGPL-3.0 for server-side/collaboration components, Apache-2.0 for the GPUI framework β verified against the repository's own LICENSE-GPL file
- Local AI model providers: Ollama, LM Studio, a local llama.cpp server, and generic OpenAI-compatible endpoints β used for the AI agent, inline assistant, and terminal AI threads
- The Personal plan (free forever) includes the full editor, multiplayer collaboration, 2,000 accepted edit predictions, and unlimited use of external AI agents or your own API keys
- Pro plan from $10/month (after a 14-day trial) adds unlimited edit predictions and $5/month of included usage for Zed-hosted cloud models, billed at API list price plus 10% beyond that
- Business plan at $30/seat/month adds org-wide AI model policies, spend visibility, and role-based access controls
- GitHub repository (github.com/zed-industries/zed) shows 90,142 stars as of September 12, 2026, per the GitHub API
- Available on macOS, Linux, and β as a fully official, stable platform since October 2025 β Windows, via a DirectX-based rendering backend
π In One Sentence
Zed is a Rust-based code editor with a built-in AI agent, inline edit predictions, and multiplayer collaboration, whose AI features can run entirely against a local model through Ollama or LM Studio instead of a cloud provider.
π¬ In Plain Terms
It works like a normal fast code editor β open a folder, edit files, run a terminal β but with an AI chat panel built in. Instead of that AI panel always calling a cloud service, you can point it at a model running on your own computer, so your code never has to leave your machine if you set it up that way.
πNote: This review is the deep-dive companion to Zed's entry in the Local LLM Software Directory β see that page for how Zed compares at a glance to dozens of other local-AI-capable tools.
What Is Zed?
Zed is a code editor written in Rust, built by Zed Industries, a company founded by former Atom and Electron engineers. Its own software overview page describes it as a high-performance, multiplayer editor. Unlike editors that bolt AI features onto an existing codebase, Zed was rebuilt from scratch around its own GPUI UI framework specifically to make GPU-accelerated rendering and low-latency AI interaction β like streaming edit predictions inline as you type β architecturally possible.
- Core function: a general-purpose code editor (syntax highlighting, multi-cursor editing, a built-in terminal, Git integration, LSP support) with an AI agent, inline "edit prediction," and an Inline Assistant layered on top
- Rendering: GPUI, a GPU-accelerated UI framework built specifically for Zed and released under Apache-2.0
- Local inference: no bundled inference engine of its own β Zed connects out to a separately running local server (Ollama, LM Studio, or llama.cpp in router mode) or a cloud API, rather than shipping model weights inside the editor
- Developer: Zed Industries, founded by Nathan Sobo and Antonio Scandurra, previously engineers on Atom and Electron at GitHub
- Canonical repository: github.com/zed-industries/zed
Zed's Background and Milestones
Zed Industries was founded by Nathan Sobo and Antonio Scandurra, who previously worked on Atom) and Tree-sitter at GitHub. The editor was built from the ground up in Rust rather than as an Electron app, specifically to target the performance ceiling the team felt Atom could not reach.
- 1January 2024 β Zed becomes fully open source
Why it matters: Zed Industries published the editor's full source code, including server-side collaboration code, moving from a closed beta to a publicly buildable, source-available project on GitHub. - 22024β2025 β AI agent and inline assistant added
Why it matters: Zed layered an AI chat agent and an Inline Assistant (for in-place code edits driven by a prompt) on top of the base editor, with support for both cloud providers and self-hosted local models from the start of the local-model feature. - 3October 2025 β Official Windows support ships
Why it matters: Windows moved from unofficial community builds to a fully supported platform with weekly updates, using a DirectX-based rendering backend, per Zed's own [Windows launch post](https://zed.dev/blog/zed-for-windows-is-here). - 42025β2026 β Local model provider list expands
Why it matters: Zed added structured support for Ollama, LM Studio, and llama.cpp in router mode as first-class local providers for the AI agent and Inline Assistant, documented in Zed's own [local-model guide](https://zed.dev/docs/ai/use-a-local-model).
What Can You Do With Zed?
Zed's feature set combines a conventional fast code editor with AI features that can run against either a cloud model or a fully local one, per Zed's own documentation and software overview.
- AI agent (Zed Agent) β a chat-and-edit agent that can read and modify files in your project, run against local models (Ollama, LM Studio, llama.cpp) or cloud providers you configure with your own API key
- Inline Assistant β an in-place editing mode triggered from the editor itself, for rewriting a selection or generating code at the cursor using whichever model backend you have configured, local or cloud
- Edit Prediction β inline, streaming next-edit suggestions as you type; this specific feature has its own setup path and is not automatically included when you configure a local-model provider for the agent
- Local model support β four documented local backends: Ollama, LM Studio, a local llama.cpp server in router mode, and any generic OpenAI-compatible endpoint, per Zed's local-model guide
- External agent support β Zed can also front external CLI agents such as Claude Code or Codex CLI inside its own interface, separate from its own built-in AI agent
- Multiplayer collaboration β real-time collaborative editing between multiple developers in the same project, a feature carried over from Zed's original design goals
- Terminal AI Threads β AI-assisted terminal sessions, configurable to use a local model provider the same way the main agent is
- Bring-your-own-key cloud access β connect your own API key from Anthropic, OpenAI, Google AI, Amazon Bedrock, GitHub Copilot, DeepSeek, Mistral, OpenRouter, and others, at no additional charge from Zed itself
Usage Examples: Three Ways to Use Zed
These are concrete workflows built from Zed's own documentation, not hypothetical use cases.
Get Zed for macOS, Windows, and Linux
Zed is a free download on macOS, Windows, and Linux. No paid plan or account is required to install and use the core editor. Links below are from the official download page β always verify against that page directly, since build URLs can change between releases.
Platform | Get It |
|---|---|
| macOS | Zed.dmg (Apple Silicon and Intel builds) |
| Windows | Installer β officially supported since October 2025 |
| Linux | One-line install script or distro package β see the download page for exact commands |
Zed has no fixed hardware minimum for the editor itself β the practical requirement is whichever local AI model you choose to load through Ollama or LM Studio, since that model's size determines your RAM/VRAM need, not Zed.
Zed Pricing: Free Editor, Optional Paid AI
The Zed editor itself is free on the Personal plan, permanently. Paid plans exist specifically for Zed-hosted cloud AI usage and organization-wide governance β neither is required to use Zed's local-model support, per Zed's own pricing page.
Personal
- Price:
- Free forever
- What It Includes:
- Full editor, multiplayer collaboration, 2,000 accepted edit predictions, and unlimited use of local models or your own API keys with external agents
Pro
- Price:
- 14-day free trial, then $10/month
- What It Includes:
- Unlimited edit predictions, plus $5/month of included usage for Zed-hosted cloud models, billed at API list price plus 10% beyond that allowance
Business
- Price:
- $30/seat/month, no seat minimum
- What It Includes:
- Org-wide AI model policies and data governance, unified spend visibility, unlimited edit predictions, and role-based access controls
Prices and plan details change; check zed.dev/pricing directly before deciding. AI features can also be disabled entirely, and bring-your-own-key cloud usage (Anthropic, OpenAI, Google AI, Amazon Bedrock, and others) carries no additional charge from Zed itself beyond what your provider bills you.
Zed vs. Cursor
Zed and Cursor are both AI-native code editors, and they get compared often because both build AI features directly into the editor rather than as an extension. The clearest differences are the underlying codebase, license, and how thoroughly each supports fully local, offline model use.
Base editor
- Zed:
- Built from scratch in Rust with a custom GPUI rendering framework
- Cursor:
- A fork of VS Code (Electron/TypeScript), see PromptQuorum's Cursor local mode review
License
- Zed:
- GPL-3.0 (editor), AGPL-3.0 (server code), Apache-2.0 (GPUI) β source-available and buildable from GitHub
- Cursor:
- Proprietary; VS Code portions remain MIT, Cursor's own additions are closed source
Local model providers
- Zed:
- Ollama, LM Studio, llama.cpp (router mode), generic OpenAI-compatible endpoints
- Cursor:
- Local models reachable mainly through a custom OpenAI-compatible API override, per PromptQuorum's dedicated review
Free tier
- Zed:
- Full editor free forever; 2,000 accepted edit predictions included
- Cursor:
- Free tier exists with limited monthly AI usage; see Cursor's own pricing for current limits
Collaboration
- Zed:
- Built-in real-time multiplayer editing
- Cursor:
- No native real-time multiplayer editing feature
If a source-available, non-Electron editor with native multiplayer and a documented local-model setup matters to you, Zed is the more direct fit. If you specifically want a VS Code-compatible extension ecosystem, evaluate Cursor directly β see PromptQuorum's Cursor local mode review for that tool's exact local-model setup and limitations.
Who Should Use Zed?
Whether Zed fits depends on whether editor performance and a documented local-model path matter more to you than an existing VS Code extension library.
Competitors and Alternatives
Zed sits in the same segment as several other AI-capable code editors and IDE plugins with local-model support. This review is the deep-dive companion to Zed's entry in the Local LLM Software Directory β see that page for the full catalog.
- Continue.dev β an open-source IDE plugin (VS Code and JetBrains) rather than a standalone editor, with similarly broad local-model provider support; see the Continue.dev review.
- Cline β a VS Code extension built around an autonomous coding agent, with local-model support through Ollama and LM Studio; see the Cline review.
- Cursor β a VS Code fork with a closed-source AI layer; see the Cursor local mode review and the head-to-head comparison above.
- Windsurf β another VS Code-fork-based AI editor competing directly with Cursor and Zed on built-in agent features; see the Windsurf review.
Common Mistakes When Evaluating Zed
Most confusion about Zed comes from treating it as either fully closed-source or fully unified under one license, or assuming Windows support is still unofficial.
Frequently Asked Questions
What is Zed?
Zed (zed.dev, source at github.com/zed-industries/zed) is a code editor built in Rust by the creators of Atom and Tree-sitter, with a built-in AI agent, inline edit predictions, and multiplayer collaboration.
Is Zed free?
The core editor (Personal plan) is free forever on macOS, Windows, and Linux, including 2,000 accepted edit predictions and unlimited use with local models or your own API keys. A paid Pro plan (from $10/month) and Business plan ($30/seat/month) add unlimited edit predictions, Zed-hosted cloud model credits, and org-wide governance β neither is required for local-model use.
Is Zed open source? What license does it use?
Zed is source-available under multiple licenses, not one: GPL-3.0 for the core editor, AGPL-3.0 for server-side/collaboration components, and Apache-2.0 for its GPUI UI framework, verified against the repository's own LICENSE-GPL file.
Does Zed support local AI models?
Yes. Zed's AI agent, Inline Assistant, external agents, and Terminal AI Threads can all run against a fully local model through Ollama, LM Studio, a local llama.cpp server in router mode, or any generic OpenAI-compatible endpoint, per Zed's local-model guide.
Does Zed run on Windows?
Yes. Windows became a fully, officially supported platform in October 2025, with weekly updates alongside macOS and Linux, using a DirectX-based rendering backend, per Zed's Windows launch post.
How many GitHub stars does Zed have?
Zed's repository (github.com/zed-industries/zed) showed 90,142 stars as of September 12, 2026, verified directly via the GitHub API. Check the repository directly for a current count, since it changes daily.
Who develops Zed?
Zed Industries, founded by Nathan Sobo and Antonio Scandurra, both previously engineers on Atom and Electron at GitHub.
Does Zed support VS Code extensions?
No. Zed has its own, smaller extension ecosystem and is not built on VS Code, so existing VS Code extensions are not directly compatible.
Does local-model support in Zed cover Edit Prediction?
Not automatically. Configuring a local provider for the AI agent and Inline Assistant is a separate step from Edit Prediction, which has its own dedicated setup per Zed's documentation.
What local model providers does Zed support?
Four documented options: Ollama, LM Studio, a local llama.cpp server in router mode, and generic OpenAI-compatible endpoints, per Zed's local-model guide.