Zed
6.6/10Zed is an open-source code editor written in Rust. It has a built-in AI agent, next-edit predictions and support for external coding agents, for developers and engineering teams.
Data last updated November 2026 · Our methodology
Overview
Zed is a native code editor built in Rust by Zed Industries, the team that created Atom and Electron. It draws with its own GPU-accelerated UI framework, GPUI, and runs on macOS, Linux and Windows. Zed reached version 1.0 in late April 2026. AI is optional and can be switched off entirely. The agent panel can edit across a codebase and shows its changes as an editable diff that you accept or reject. Parallel agents can work on different tasks at once. Zeta, Zed's open-weight model, predicts your next edit as you type. Through the Agent Client Protocol (ACP), which Zed created, Claude Agent, Codex CLI, OpenCode and other external agents run inside the editor. MCP servers can extend the agent's tools. Pricing is simple. The free Personal plan works with your own API keys, local models through Ollama, or external agents. Pro costs $10 a month and includes unlimited edit predictions plus $5 of hosted model usage. Beyond that, usage is billed at provider list price plus 10%. Business costs $30 per seat per month with no bundled AI credit. It adds org-wide model and data-sharing policies, per-member spend visibility and role-based access. Bring-your-own-key providers include Anthropic, OpenAI, Google, Amazon Bedrock, GitHub Copilot, Mistral, DeepSeek and OpenRouter. Zed suits developers who care about speed, Linux users, and teams that want free choice of AI providers or fully local models. The main limitations are a much smaller extension ecosystem than VS Code and a debugger that is weaker for some languages. Larger companies should also note that SSO, SAML, SCIM and SOC 2 were not yet available when this was checked.
Our Rating
Who Zed is best for
Best software for DevelopersGood fit
Zed is a code editor built for programmers, with agent, edit prediction and debugging features.
The free Personal plan with bring-your-own-key or local models costs nothing beyond model usage.
The free tier, low-cost Pro plan and month-to-month Business seats with no minimum suit small engineering teams.
Built-in multiplayer editing, channels and voice chat support remote pair programming.
Less suited
Business adds org-wide AI policies and spend tracking, but extension gaps can slow adoption.
SSO, SAML, SCIM and SOC 2 are not yet available, which blocks many enterprise rollouts.
Pros & Cons
Pros
- Very fast native app: one reviewer measured a 0.4 s cold start and 180 MB of idle RAM, against 3.0 s and 650 MB for VS Code with Copilot.
- No lock-in to one AI provider. The free plan works with your own keys from many providers, local Ollama models, or ACP agents such as Claude Agent and Codex.
- Cheap paid tiers: Pro is $10 a month, and hosted model usage is marked up only 10% over list price, with an optional spending cap.
- The editor, the Zeta edit-prediction model and ACP are all open source, and AI can be switched off entirely.
- Detailed agent controls: per-tool permissions and MCP profiles, plus agent changes shown as a reviewable diff.
- Privacy defaults suit companies: Zed keeps no prompts by default, hosted providers must apply zero data retention apart from some designated models, and data sharing is off by default on Business.
Cons
- About 1,000 extensions against more than 50,000 for VS Code, and users say most are language servers and themes.
- The debugger is less mature for some languages, such as .NET and Java.
- SSO, SAML, SCIM and SOC 2 were not yet available, which limits use in regulated companies and enterprises.
- Reviewers found rough edges in the agent experience: no built-in memory or skills feature, and inconsistent support across external agents.
- Settings and keymaps are edited as JSON, and there is no VS Code import wizard, so switching takes effort.
- Business seats include no AI credit, so hosted model costs come on top of the $30 seat price.
Key Features
Zed Pricing
Personal
- Free forever
- Local models through Ollama
- Unlimited use with your own API keys
- AI features can be turned off entirely
- External agents such as Claude Agent and Codex CLI
Pro
- Optional spending cap
- Unlimited edit predictions
- 14-day free trial with no credit card
- Overage billed at API list price plus 10%
- $5 of hosted model tokens included each month
Business
- Role-based access controls
- Unlimited edit predictions
- Unified spend visibility per member
- No minimum seats, month-to-month billing
- Invoiced order forms for 25+ seats via sales
- Org-wide AI model policies and data governance
- Admins can disable hosted models, edit predictions and collaboration
- No bundled AI credits; hosted AI at usage rates or your own API keys
Frequently Asked Questions
Yes. Pro has a 14-day trial with no credit card. It includes $5 of usage on a single hosted model and unlimited edit predictions, and ends when either the credit or the 14 days runs out. Business had no free trial when this was checked.
Yes. The free Personal plan allows unlimited use with your own API keys, local models through Ollama, or external agents such as Claude Agent and Codex CLI.
Not yet. Zed says SSO, SAML, SCIM and SOC 2 are not currently available but are in development, and that it is working toward SOC 2 Type 1. Compliance questions go to sales@zed.dev.
Zed says it keeps no prompts or code context by default, and its agreements with hosted providers forbid training on them. Edit-prediction training data is collected only if you opt in, the project is open source, and the file is not excluded. Some designated Anthropic models keep data for 30 days for safety review.
Business costs $30 per seat per month, billed month to month, with no seat minimum. Teams of 25 or more seats can arrange invoiced order forms through sales. Hosted AI is billed at provider cost plus 10%. With your own keys, you pay your provider directly.
Zed runs natively on macOS, Linux (X11 and Wayland) and Windows. Through the Agent Client Protocol, it can run external agents such as Claude Agent, Codex CLI, Gemini CLI and OpenCode alongside its own agent.
Similar Tools
See all alternatives
Claude Code is Anthropic's agentic coding tool. It reads a codebase, edits files, runs commands and tests, and opens pull requests for developers and engineering teams.
GitHub Copilot is an AI coding assistant from GitHub that provides code completions, chat, code review and autonomous coding agents for individual developers and engineering organizations.
Agentic AI code editor, formerly Windsurf and now sold by Cognition as Devin Desktop, for developers and engineering teams who code with local and cloud AI agents.
Ranked near Zed in AI Software
See Zed in the listZed is #151 in AI Software by our score.
