# Zed

> A fast, Rust-built code editor for coding with teammates and AI agents

- Category: [Coding AI](https://finderatlas.com/categories/coding-ai)
- Pricing: Freemium
- Website: https://zed.dev
- Platforms: macOS, Windows, Linux, CLI
- Made by: Zed Industries, founded 2021
- Last updated: 2026-10-19
- FinderAtlas page: https://finderatlas.com/tools/zed

## Overview

Zed is a code editor written in Rust and built for speed and for collaboration between developers and AI. It runs natively on macOS, Linux and Windows, using multiple CPU cores and the GPU to keep editing responsive. The editor is developed by Zed Industries, a team led by Nathan Sobo, Antonio Scandurra and Max Brunsfeld, who previously worked on Atom, Electron and Tree-sitter. On the AI side, Zed includes its own agent, an Agent Panel, parallel agents, an inline assistant and edit prediction powered by the open-weight Zeta2 model. It can also run external agents such as Claude Agent, Codex and OpenCode through the Agent Client Protocol (ACP), and connect to MCP servers. Developers can use Zed-hosted models, their own API keys, provider subscriptions or local models.

## Key features

- Rust-Native Performance
- Agentic Editing
- External Agents via ACP
- Edit Prediction
- Multiplayer Collaboration
- Debugger
- Native Git Support
- Remote Development
- Flexible Model Access

## Pricing

- **Personal** ($0): Free editor with your own API keys or external agents
- **Pro** ($10/month): Individuals who want Zed-hosted models and edit predictions
- **Business** ($30/user/mo): Teams that need AI policy and data controls

## Best for

- Fast native editing, running AI agents alongside a conventional editor, real-time pair programming

## Who uses it

- Software Engineers
- Development Teams
- Open-Source Contributors
- Developers who prefer lightweight native editors

## Strengths

- Native Rust application designed for low-latency editing on multiple cores and the GPU.
- Free Personal plan with unlimited use of your own API keys or external agents.
- Supports external agents through ACP and MCP rather than locking users to one agent.
- Built-in multiplayer collaboration, debugger, Git and remote development.
- Source code is public on GitHub.

## Limitations

- No web or mobile version; the README lists web as not yet available.
- Hosted AI requires a paid plan, billed at API list price plus 10%.
- Business plan SSO, SAML and SCIM are planned but not yet available.
- Extension library is described on the site as hundreds of extensions, so niche tooling may need workarounds.

## Not recommended for

- Developers who need a browser or mobile editor, teams that need SSO or SAML today

## Integrations

- Claude Agent
- Codex
- OpenCode
- MCP servers
- Language Server Protocol
- Debug Adapter Protocol
- Jupyter
- Git
- GitHub

## FinderAtlas editorial scores (out of 10)

- Ease of Use: 8.2/10
- Feature Depth: 8.4/10
- Value for Money: 8.6/10
- Performance & Speed: 9.3/10
- Documentation & Support: 8/10

## Alternatives

- [Augment Code](https://finderatlas.com/tools/augment-code): Agentic coding platform that runs code review and ticket-to-PR work on your codebase ([compare](https://finderatlas.com/compare/augment-code-vs-zed))
- [Amp](https://finderatlas.com/tools/amp): Coding agent that runs in cloud orbs and keeps working after you close your laptop ([compare](https://finderatlas.com/compare/amp-vs-zed))
- [OpenAI Codex](https://finderatlas.com/tools/openai-codex): OpenAI's coding agent that writes, reviews and runs code from terminal, IDE, desktop or cloud ([compare](https://finderatlas.com/compare/openai-codex-vs-zed))
- [Kiro](https://finderatlas.com/tools/kiro): Spec-driven agentic IDE from AWS that turns prompts into designs, tasks and code ([compare](https://finderatlas.com/compare/kiro-vs-zed))
- [Macroscope](https://finderatlas.com/tools/macroscope): AI code review, agent orchestration and engineering status updates for GitHub teams ([compare](https://finderatlas.com/compare/macroscope-vs-zed))
- [Kilo Code](https://finderatlas.com/tools/kilo-code): Open-source AI coding agent for VS Code, JetBrains and CLI with 500+ models ([compare](https://finderatlas.com/compare/kilo-code-vs-zed))
- [Google Antigravity](https://finderatlas.com/tools/google-antigravity): Google's agent-first platform for building software with parallel AI coding agents ([compare](https://finderatlas.com/compare/google-antigravity-vs-zed))
- [Cursor](https://finderatlas.com/tools/cursor): AI coding agent and VS Code-based editor for multi-file edits and cloud agents ([compare](https://finderatlas.com/compare/cursor-vs-zed))

## Recent updates

- 2026-10-01 (v1.24.3): Adds Gemini 4 Argon to the Google AI models
- 2026-10-01 (v1.24.2): Adds Git tag creation, lets you drag Terminal tabs into the Agent Panel, and pre-fills File Finder search with select...
- 2026-10-01 (v1.23.3): Patch release fixing Markdown preview fonts and edit predictions for single-worktree files, and updating to Node.js v...

## FAQ

**Is Zed free?**

Yes. The Personal plan is free forever and works with your own API keys or external agents, but does not include Zed-hosted AI. Pro costs $10 per month and Business $30 per seat per month.

**Which operating systems does Zed support?**

Zed is available on macOS, Linux and Windows. The project README lists a web version as not yet available.

**Does Zed train on my code?**

According to Zed's documentation, it does not retain prompts or code context by default and has no-training commitments from model providers for Zed-hosted models. Edit prediction training data is collected only if you opt in, the project is open source and the file is not excluded.

**Can I use other AI agents in Zed?**

Yes. Zed runs external agents such as Claude Agent, Codex and OpenCode through ACP, and can connect to MCP servers. Agent CLIs can also run in terminal threads.

**How are hosted AI models billed?**

Hosted usage is billed at API list price plus 10%. Pro includes $5 of token credit, and further usage is billed at month-end or per additional $10, whichever comes first.

**Can I turn AI off?**

Yes. The documentation includes a page on turning AI off, and Business admins can disable edit predictions organization-wide.