Zed Editor Review 2026: AI Agents, Pricing & Limits

Disclosure: CodeToolPilot may earn a commission if you buy through links on this page, at no extra cost to you. How we make money.

Last verified: October 12, 2026. This is an editorial research review — how we review. Prices and plans change often; confirm on the vendor’s site.

Zed is a code editor written from scratch in Rust by Zed Industries, the team behind the Atom editor and the Tree-sitter parsing library. It is not a fork of VS Code. It draws its interface on the GPU with its own UI framework, and its pitch has always been speed and real-time collaboration. Zed reached version 1.0 in April 2026 and still ships weekly releases for macOS, Linux and Windows.

This review focuses on Zed’s AI features, which are built in but optional. Zed is for developers who want a fast, native editor and want to choose their own AI setup. You can use Zed’s own agent, plug in Claude Agent, Codex, Gemini CLI or other agents, bring your own API keys, use an existing subscription, or run local models. You can also turn AI off entirely.

This is an editorial research review based on Zed’s pricing page, documentation, blog and release notes. We did not run our own benchmarks or timed tests.

Quick verdict

Of the editors in this category, Zed is the most flexible and the least tied to one vendor. It is fast and open source, the free tier is generous for anyone with their own keys or agent subscriptions, and Pro costs only $10 a month. In exchange, it does less out of the box than Cursor or Devin Desktop for cloud agents and automated PR review, VS Code extensions do not carry over, and the Business plan still lacks SSO and SOC 2.

Pro: A native Rust editor focused on performance, with an open-source core.
Pro: Bring your own agent: Zed Agent plus external agents over the Agent Client Protocol (ACP), such as Claude Agent, Codex, Gemini CLI, OpenCode, Copilot and Cursor.
Pro: A useful free tier: unlimited use with your own API keys or external agents, plus 2,000 accepted edit predictions a month.
Pro: Privacy-minded defaults. Zed says it does not retain prompts or code by default, and hosted model providers are bound to no-training and zero-data-retention terms.
Con: No VS Code extension compatibility. Zed has its own extension system.
Con: Zed for Business does not yet offer SSO, SAML, SCIM or SOC 2.
Con: Pro includes only $5 of hosted tokens a month. Beyond that, usage is billed at list price plus 10%.
Con: Fewer built-in cloud agent and review features. Zed’s agent collaboration work is in Delta, a separate product in public beta.

Visit Zed

What is Zed?

Zed is a general-purpose editor with language server support, Git integration, a debugger, SSH remote development, a terminal and built-in multiplayer collaboration. Its 1.0 announcement describes it as “an AI-native editor”: AI is part of the foundation but can be switched off. The latest stable release at the time of writing is 1.23.2, published October 7, 2026.

Zed organises AI around three ways to run agents. Zed Agent is the built-in agent and uses whatever models you configure. External Agents run in their own process through ACP and handle their own login and billing. Terminal Threads run an agent CLI or TUI inside a terminal-backed thread. A Threads Sidebar, added in April 2026, lets you run many of these in parallel across projects. In August 2026 the company also introduced Delta, a separate multiplayer environment for coding with agents and reviewing their work without pull requests. Delta is free during its public beta.

Key features

Agent Panel and Zed Agent

The Agent Panel is where you prompt agents, add context, review changes and manage threads. Zed Agent includes built-in tools for editing files, searching code, running terminal commands and web search. It also supports profiles, MCP servers, skills, AGENTS.md-style instructions and checkpoints. Agent tool calls can run in an OS-level sandbox.

External Agents and Terminal Threads

Zed created the Agent Client Protocol in 2025 and runs an ACP Registry. From the registry you can install Claude Agent, Codex, Gemini CLI, OpenCode, Copilot, Cursor and other agents, each of which keeps its own authentication and billing. Zed does not charge for external agents. Terminal Threads let you run an agent’s own CLI or TUI directly inside Zed.

Parallel Agents

The Threads Sidebar groups agent threads by project, lets you use a different agent in each thread, controls which folders and repositories each agent can access, and can isolate work in Git worktrees. For existing users, the new agent-first panel layout is opt-in.

Edit Prediction

Edit Prediction offers multi-line suggestions as you type. The default provider is Zeta, an open model developed by Zed. GitHub Copilot, Codestral and Mercury Coder are alternatives. The free plan includes 2,000 accepted Zeta predictions a month, and Pro removes the limit.

Flexible model access

Zed can use Zed-hosted models, your own API keys or gateways, and existing subscriptions such as ChatGPT, Claude or Copilot. Supported providers include Amazon Bedrock, Anthropic, Google AI, OpenAI, OpenRouter, Mistral and DeepSeek. Local models work through Ollama, LM Studio or llama.cpp. Zed’s hosted list includes current Claude, GPT and Gemini models.

Collaboration

Built-in channels, shared projects, screen sharing and following mode let teammates edit together in real time. You can also follow an agent as it navigates and edits.

Who should use Zed

Zed suits developers who value editor speed and want to choose their own AI tools. That includes people who already pay for Claude, ChatGPT or Copilot and want to use those subscriptions in a fast editor, and teams who want AI that can be tightly limited or switched off. It also suits pair programmers who like built-in real-time collaboration.

Who should skip it

Skip Zed if your company requires SSO or SOC 2 before approving tools, if you rely on VS Code-only extensions, or if you want a fully managed agent platform with cloud agents and automated PR review in one subscription.

Strengths and limitations

Strengths

Performance focus. Zed’s native Rust and GPU architecture is its defining design choice. The company built its own UI framework rather than using Electron.
Freedom of choice. Few editors make it this easy to mix Zed’s own agent, third-party agents, local models and your own keys, all without a subscription.
Low and controllable cost. A $10 Pro plan with token pass-through pricing and a hard spend limit is easy to budget.
Privacy controls. No retention by default, opt-in training data only, and Business admins can lock data-sharing settings so individuals cannot override them.
Open source. The editor’s code is public, which matters to teams that want to audit or contribute.

Limitations

Extension ecosystem. VS Code extensions do not work in Zed. If you depend on niche VS Code tooling, check Zed’s extension gallery first.
Enterprise gaps. Zed says SSO, SAML, SCIM and SOC 2 are not yet available but are in progress. Many security reviews require them.
Thin included credit. At Zed’s listed rate for Claude Opus 5.5 ($22 per million output tokens), $5 of credit does not go far for heavy agent use, so most Pro users will either pay usage or bring their own keys.
Fewer managed agent services. Zed’s plans do not include a managed cloud-agent service or bot-driven PR review like those bundled by Cursor or Cognition. Delta covers some team and agent collaboration needs, but it is a separate beta product.
External agents have their own terms. Billing, data retention and legal terms for ACP agents sit with each provider, not with Zed.

Pricing (as of October 2026)

Plan	Price	What you get
Personal	$0	The full editor, 2,000 accepted edit predictions a month, and unlimited use with your own API keys or external agents. No Zed-hosted models.
Pro	$10/month	Unlimited edit predictions, all Zed-hosted models and $5 of token credit each month. Further usage is billed at API list price plus 10%, with a spend limit you control.
Business	$30/seat/month	Hosted models and unlimited edit predictions for members, org-wide AI model policies, enforced data-sharing settings, role-based access and consolidated billing. No bundled AI credits.
Student	Free for one year	For verified university students: unlimited edit predictions and $10 a month in token credits, with the most expensive models excluded.

Pro comes with a 14-day trial that needs no credit card and includes $5 of GPT-6 Luna usage. Pro’s default spend limit caps extra token usage at $10 a month, so by default the most you will pay is $20 a month before tax unless you raise the limit. Business is month-to-month with no seat minimum, and order forms are available from 25 seats. Pricing changes regularly, so confirm the current details on Zed’s pricing page. See Zed pricing

Prices verified on October 12, 2026 against the official Zed pricing page. Plans change often — confirm on the vendor’s page before you buy.

Use cases

Zed’s AI features are optional and configurable, so its best use cases depend on how much of your own AI setup you want to bring. These scenarios are based on Zed’s documentation and the research behind our review, not on hands-on testing.

Using an existing AI subscription in a fast native editor

A developer who already pays for Claude, ChatGPT or Copilot wants an agent in a fast editor without buying another AI plan. Zed can run Claude Agent, Codex, Copilot and other agents as external agents over the Agent Client Protocol, and it can also use existing ChatGPT, Claude or Copilot subscriptions as model sources. Zed does not charge for external agents. The caveat is that each external agent keeps its own billing, data retention and legal terms.

Working with local models or with AI switched off

A developer working on code that should not leave the machine wants AI help from local models, or no AI at all. Zed connects to local models through Ollama, LM Studio or llama.cpp, and its AI features can be turned off entirely. This makes it a reasonable fit for privacy-sensitive work, though the quality of help then depends on the local model and hardware you run.

Running several agents in parallel

A developer juggling several tasks across projects wants to hand each one to a separate agent and review the results. Zed’s Threads Sidebar groups agent threads by project, allows a different agent in each thread, controls which folders and repositories each agent can access, and can isolate work in Git worktrees. Existing users need to opt in to the newer agent-first panel layout.

Real-time pair programming

A small distributed team wants to edit the same code together and follow each other, or an agent, as changes are made. Zed has built-in channels, shared projects, screen sharing and following mode, and you can follow an agent as it navigates and edits. In our assessment, this works best when the whole team uses Zed, since these features are built into the editor itself. For multiplayer coding with agents and reviewing their work without pull requests, Zed’s separate Delta product is in public beta.

A small team that wants central AI policy

A startup engineering team wants consistent AI rules without signing an enterprise contract. Zed for Business adds org-wide AI model policies, enforced data-sharing settings, role-based access and consolidated billing, and it is month-to-month with no seat minimum. The caveat is that SSO, SAML, SCIM and SOC 2 are not yet available, and many security reviews require them.

Alternatives

Cursor: a VS Code-style AI editor with cloud agents, Bugbot reviews and enterprise SSO.
Devin Desktop (formerly Windsurf): a VS Code-compatible IDE that also runs ACP agents and hands work off to Devin’s cloud agent.
Claude Code: a terminal-based coding agent that can also run inside Zed as an external agent via Claude Agent.
GitHub Copilot: an assistant that works across many editors and can also act as Zed’s edit-prediction provider.

Final verdict

Zed is a fast, open editor that treats AI as something you configure, not something you are locked into. Its free tier and $10 Pro plan are some of the most affordable ways to get serious AI help, and ACP support lets you bring the agents you already use. Teams that need SSO, compliance certifications or managed cloud agents should look elsewhere for now. Everyone else can try it free.

Download Zed

FAQ

Is Zed free?

Yes. The Personal plan is free and includes the full editor, a monthly allowance of accepted edit predictions and unlimited use with your own API keys or external agents, but no Zed-hosted models. Paid plans add Zed-hosted models and unlimited edit predictions, and the editor’s code is open source either way. Check Zed’s pricing page for current allowances.

Can I use Claude Code or OpenAI Codex inside Zed?

Yes, through the Agent Client Protocol (ACP), which Zed created in 2025. You can install Claude Agent, Codex, Gemini CLI, OpenCode, Copilot, Cursor and other agents from Zed’s ACP Registry, or run an agent’s own CLI in a Terminal Thread. Each external agent keeps its own login and billing, and Zed does not charge for them. Our Claude Code and OpenAI Codex reviews cover those agents in more detail.

Can Zed use local AI models?

Yes. Zed supports local models through Ollama, LM Studio and llama.cpp, alongside your own API keys for providers such as Anthropic, OpenAI, Google AI, Amazon Bedrock, OpenRouter, Mistral and DeepSeek. You can also switch Zed’s AI features off entirely.

Does Zed support VS Code extensions?

No. Zed is written from scratch in Rust rather than forked from VS Code, and it has its own extension system. If you depend on specific VS Code tooling, check Zed’s extension gallery before switching.

Is my code used to train AI models in Zed?

Zed says it does not retain prompts or code by default, that sharing data for training is opt-in, and that its hosted model providers are bound to no-training and zero-data-retention terms. On the Business plan, admins can enforce data-sharing settings so individuals cannot override them. External agents run under their own providers’ data and legal terms, so check those separately.

Zed vs Cursor: which is better?

It depends on what you want bundled. Cursor is a VS Code-style AI editor with cloud agents, Bugbot reviews and enterprise SSO, while Zed is a native Rust editor that lets you bring your own agents, models and keys and has a generous free tier. In our assessment, Zed suits developers who prioritize speed, openness and low cost, and Cursor suits teams that want managed cloud agents, automated PR review and SSO in one subscription.

Zed Review (2026): A Fast, Open-Source Editor With Bring-Your-Own AI