Zed Dev Stack
BeginnerA fast Rust editor with its own agent, and room to add Claude Code, Codex, or others.
Published 12 October 2026
About Zed Dev Stack
Zed is an open-source code editor written in Rust, drawn on the GPU rather than through a browser engine, so it opens instantly and stays responsive on large files. It comes with its own agent, the Zed Agent, which edits across files, runs commands, calls MCP servers, and hands parts of a task to subagents. Several agent threads can run at once against different projects or git worktrees, and edit prediction suggests the next change as you type. Collaboration is built in too: shared projects, screen sharing, and editing together with live cursors.
The Zed Agent runs whichever model you pick. Claude is the default here; GPT and Gemini sit beside it as Zed-hosted models on the Pro plan, billed at the provider's list price plus 10%, and DeepSeek, Grok, and Mistral connect with your own API keys, as can Claude and OpenAI. A ChatGPT Plus or Pro sign-in also works as a provider, and the agent can run a local model when one is added. Zed charges nothing for the editor or for external agents; its paid plans add hosted models, unlimited edit predictions, and team controls.
Zed is also an agent host. You can add Claude Code, Codex, GitHub Copilot, OpenCode, Grok Build, or dozens of other agents from a registry inside the editor; they run over the Agent Client Protocol, with their changes shown in Zed's diff review, while Terminal Threads run any command-line agent as a managed thread in the sidebar. Each agent keeps its own subscription and model.
Zed runs on your own computer, which has to stay on: it has no hosted cloud agents and no phone app. Its built-in remote development opens a project on a server over SSH, with code, terminals, and language servers running there and the interface on your laptop, which also makes the Zed Agent's model calls, so the laptop still has to be open. For work that continues with it closed, an added agent's own cloud, such as Claude Code on the web or Codex cloud tasks, runs without Zed.
Key Features
- ✓A native Rust editor with GPU rendering, Vim mode, and built-in collaboration
- ✓The Zed Agent with subagents and MCP, on Claude by default or GPT, Gemini, DeepSeek, Grok, or Mistral
- ✓Zed-hosted models on the Pro plan, or your own API keys and a ChatGPT sign-in on the free plan
- ✓Optional external agents such as Claude Code, Codex, and Copilot over the Agent Client Protocol, from an in-editor registry
- ✓Terminal Threads that run any command-line agent as a managed sidebar thread
- ✓Parallel agent threads across projects and git worktrees
- ✓SSH remote development, with code, terminals, and language servers on a server
When to Use Zed Dev Stack
- →Adding Claude Code, Codex, or Copilot and trying them on the same task without switching editors
- →Keeping an editor fast on a large codebase while an agent works in the sidebar
- →Running the Zed Agent on your own API keys or a ChatGPT plan instead of another subscription
- →Editing a project that lives on a server, with builds and tests running there
- →Pairing with a teammate in a shared project while an agent drafts the next change
Pros
- Fast startup and low memory use next to Electron-based editors
- Many vendors' agents in one editor, with no lock-in to one agent or model
- Open source, and free to use with your own keys or external agents
- Collaboration and remote development are built in rather than added through extensions
Cons
- No hosted cloud agents or phone access; the laptop has to stay on while Zed's agent works
- A smaller extension ecosystem than VS Code, and a different keymap and settings to learn
- Each external agent needs its own subscription, and Claude over the Agent Client Protocol bills at API rates
- Zed's profiles and skills apply to the Zed Agent only, not to external agent threads
LLM Options for Zed Dev Stack
Zed Dev Stack powered by Claude
Claude is the default here for the Zed Agent, a common pick for large multi-file changes. As a Zed-hosted model on Pro, Sonnet 5.5 costs $2.20 in and $11 out per million tokens and Opus 5.5 $4.40 and $22; an Anthropic API key bills Anthropic's own rates. A Claude Pro or Max plan reaches Zed only through Claude Code.
Zed Dev Stack powered by OpenAI
OpenAI's GPT models are Zed-hosted on Pro, and the trial runs GPT-6 Luna at $0.11 in and $0.55 out per million tokens, a cheap model for focused edits. A ChatGPT Plus or Pro sign-in also works as a provider for the Zed Agent, with no API key, or an OpenAI key bills per token.
Zed Dev Stack powered by Google Gemini
Gemini is the third Zed-hosted family, and its long context suits questions that span a whole repository. On your own key, Gemini 3.1 Pro costs $2 in and $12 out per million tokens up to 200K, and the Flash models much less for everyday edits.
Zed Dev Stack powered by DeepSeek
DeepSeek connects to the Zed Agent with a DeepSeek API key, and Zed's own guidance names it as a cheaper route for Claude users. V4.1-Flash costs $0.15 in and $0.60 out per million tokens off-peak, double at weekday peak hours, which makes long agent sessions cheap.
Zed Dev Stack powered by Grok
Grok connects through an xAI API key, billed per token: Grok 4.7 costs $2 in and $6 out per million tokens with a 500K context. The same key pairs with Grok Build if it runs as an external agent, though a SuperGrok plan only covers Grok Build, not the Zed Agent.
Zed Dev Stack powered by Mistral
Mistral connects with a Mistral API key: Medium 3.5, tuned for agentic and coding work, costs $1.50 in and $7.50 out per million tokens, and Small 4 $0.15 and $0.60. Its free Experiment tier is enough to try it, and the weights are open for self-hosting.
These are highlighted picks. To see all the tools, check the LLM category.
Version Control Options for Zed Dev Stack
Zed Dev Stack with GitHub
GitHub is the default here, where most repositories, AI reviewers, and the agents' own cloud features live: Codex cloud tasks and Copilot's cloud agent both work from GitHub. Zed's Git panel stages, commits, and pushes to it like any remote.
Zed Dev Stack with GitLab
GitLab bundles source control, merge requests, and CI/CD in one platform you can self-host. Zed and the agents it hosts only need git, so a GitLab remote works the same in the editor; some agents' cloud features are GitHub-first.
Zed Dev Stack Add-ons
Each addition below extends this stack with a capability the base stack works fine without. None are required: include the ones your product actually needs when building this stack, and skip the rest.
Developer workflow
4
Where the agent runs, how you reach it, and the tools you work in alongside it.
Server Add-ons
Add a server when you want the agent to keep running somewhere other than your own laptop, reachable at any time and not tied to your machine staying on. Some agents also offer their own managed cloud sessions as an alternative to self-hosting; check the stack's own description for details.
Zed Dev Stack with Hetzner
Hetzner is the default here: CX23 has 4 GB of RAM for about €5.49/mo, room for Zed's remote server plus the project's language servers, builds, and a command-line agent. Zed documents no memory minimum; heavy language servers want more. A plain VPS billed hourly up to a monthly cap.
Zed Dev Stack with DigitalOcean
DigitalOcean's 4 GB Droplet is $24/mo, several times Hetzner's price for the same memory, in exchange for simple snapshots, many regions, and 1-Click images with agents such as Codex CLI and OpenCode preinstalled for a Terminal Thread to use.
Zed Dev Stack with Hostinger
Hostinger's KVM 1 has 4 GB of RAM at $6.49/mo introductory, renewing at $11.99/mo, enough for Zed's remote server and language servers on a mid-sized project. Its Claude Code template preinstalls the CLI for a Terminal Thread on that server.
Zed Dev Stack with Raspberry Pi
A Raspberry Pi 5 is a one-time hardware cost instead of a monthly bill. Zed's remote server supports Linux on arm64, so a 4 GB board hosts a small project, while 8 GB leaves room for heavier language servers and builds.
These are highlighted picks. To see all the tools, check the Hosting & Cloud category.
Remote Access Add-ons
Add remote access when you want to reach an agent running on another machine — a VPS or a home server — without exposing it to anyone but you.
Zed Dev Stack with Tailscale
Add Tailscale alongside a server so Zed's remote development reaches it over SSH on a private address from any network, with no port open to the internet. The personal tier is free.
Coding Agent Add-ons
Add a separate coding agent when you want another vendor's agent working alongside the one built into the editor, either through its own extension or in the integrated terminal.
Zed Dev Stack with Claude Code
Add Claude Code when you want Anthropic's agent beside the Zed Agent. It runs two ways: as Claude Agent over the Agent Client Protocol, which bills Anthropic's separate Agent SDK credits at API rates, or as the claude CLI in a Terminal Thread, which keeps a Claude plan's own limits. Plans start at $20/mo.
Zed Dev Stack with Codex
Add Codex when you want OpenAI's agent in the editor: it installs from Zed's agent registry and signs in with a ChatGPT plan, or runs as the codex CLI in a Terminal Thread. Codex cloud tasks keep working on GitHub after Zed is closed.
Zed Dev Stack with GitHub Copilot
Add GitHub Copilot when the team already pays for it: the Copilot agent runs over the Agent Client Protocol, and the same sign-in serves as a model provider and an edit prediction source. Copilot has a free tier and Pro at $10/mo, and its cloud agent works on GitHub issues without the editor.
Zed Dev Stack with OpenCode
Add OpenCode when you want an open-source agent that runs any provider's model on your own key, a local model included: it installs from the registry, and Zed also takes an OpenCode Zen or Go key as a model provider for the Zed Agent.
Zed Dev Stack with Grok Build
Add Grok Build when you have a SuperGrok plan or an xAI key: SpaceXAI's open-source agent runs in Zed over the Agent Client Protocol and fans larger tasks out to parallel subagents, each optionally in its own git worktree.
These are highlighted picks. To see all the tools, check the AI Coding Agents category.
Code Review Add-ons
Add code review when you want an AI reading every pull request before it merges: it comments on the diff so bugs, security issues, and inconsistencies surface before a human has to catch them.
Zed Dev Stack with CodeRabbit
Add CodeRabbit when pull requests from several agents should get one consistent reviewer: it installs as a GitHub or GitLab app, summarizes the diff, and leaves line-level comments, with a free tier for open-source repos.
Zed Dev Stack with Greptile
Add Greptile when review should read the whole repository rather than only the diff: it indexes the codebase so its GitHub and GitLab comments carry wider context, and a self-hosted deployment keeps the code on your own infrastructure.
Zed Dev Stack with Cursor Bugbot
Add Cursor Bugbot when a bug-focused reviewer is enough: it reviews GitHub or GitLab pull requests whichever agent wrote them, though it bills through a Cursor plan rather than a subscription of its own.
AI
1
Models, agents, and the pieces around them: retrieval, sandboxes, and model access.
Model Inference Add-ons
Add model inference when you want an open-weight model in the mix: on your own hardware for privacy, or on a hosted inference provider for speed and low per-token prices.
Zed Dev Stack with Ollama
Add Ollama to run the Zed Agent on an open model on your own machine: Zed's local-model docs list it as a provider. Raise Ollama's small default context window for agent work, and expect a coding model to want about 24 GB of graphics memory. The Local LLM Dev Stack covers the hardware and the models in full.
Zed Dev Stack with LM Studio
Add LM Studio for the same job from a desktop app: it downloads a model build sized for your memory and serves it on a local address that Zed connects to as a provider. On Apple Silicon it runs models through Apple's MLX engine. Free for personal and work use.
Zed Dev Stack with llama.cpp
Add llama.cpp when you want the engine without a manager around it: Zed's local-model docs cover its server, and every setting is exposed, from the quantization to the context size. The choice for full control, with the models to manage yourself. Free, MIT licensed.
These are highlighted picks. To see all the tools, check the AI Runtime & Serving category.
Build and deploy
2
Testing, packaging, and shipping your code on every change.
CI/CD Add-ons
Add CI/CD when you want a dedicated pipeline for running tests, linting, or multi-stage builds before a deploy goes out. Many hosting platforms already redeploy automatically on every push on their own — a CI/CD tool adds the most value on top of that by gating the deploy on a passing test suite, and matters even more when the hosting choice does not auto-deploy at all, such as a self-hosted server.
Zed Dev Stack with GitHub Actions
Add GitHub Actions when every push should run the tests, so changes from the Zed Agent and the external agents meet the same checks before they merge.
Zed Dev Stack with GitLab CI/CD
Add GitLab CI/CD when the project lives on GitLab: pipelines defined in a .gitlab-ci.yml file run the tests on every merge request, on GitLab's runners or your own.
These are highlighted picks. To see all the tools, check the CI/CD Pipelines category.
Containerization Add-ons
Add containerization when you want the app packaged the same way across local development, staging, and production, or need to deploy somewhere that isn't a managed serverless platform.
Zed Dev Stack with Docker
Add Docker when the project needs the same environment on a laptop, a remote server, and in CI, or when an agent left running should be fenced into a container. The agents can write the Dockerfile and Compose file when asked.
These are highlighted picks. To see all the tools, check the Containerization category.
Frequently Asked Questions about Zed Dev Stack
Can Zed keep working when my computer is off?
Not on its own. Zed has no hosted cloud agents and no phone app, and the Zed Agent's model calls come from the editor on your computer, so it stops when the laptop sleeps. Remote development puts the code, terminals, and language servers on a server over SSH, and a dropped connection reconnects to the same session, but the interface still runs locally. For work that continues with the laptop closed, use an agent's own cloud: Claude Code on the web, Codex cloud tasks, or Copilot's cloud agent on GitHub, then review the pull request in Zed.
Which agents run which models?
The model picker belongs to the Zed Agent: it runs Zed-hosted Claude, GPT, or Gemini on Pro, or Claude, OpenAI, Gemini, DeepSeek, Grok, Mistral, or a local model on your own keys. External agents bring their own: Claude Code runs Claude on a Claude plan or Agent SDK credits, Codex runs OpenAI models on a ChatGPT plan, Copilot runs the models its plan offers, Grok Build runs Grok on SuperGrok or an xAI key and can add others in its own config, and OpenCode runs whatever provider you give it. Picking Grok here doesn't change what Grok Build runs, and the reverse.
How is Zed different from VS Code with an AI extension?
Both host several vendors' agents. VS Code does it through extensions, such as Copilot, Cline, or the Claude Code and Codex extensions, in the editor with the largest extension ecosystem. Zed builds it into the editor: its own agent with a model picker, external agents over the Agent Client Protocol with a shared diff review, and Terminal Threads for anything else. Zed is faster and lighter, VS Code has more extensions and GitHub Codespaces for a browser-based editor.
Should Claude Code run over ACP or in a Terminal Thread?
It changes the bill. Since Anthropic's June 2026 change, Claude Agent over the Agent Client Protocol draws on separate Agent SDK credits ($20 a month on Pro, $100 or $200 on Max) at API rates, which run out much faster than plan limits under heavy use. The claude CLI in a Terminal Thread counts against the Claude plan's normal limits. ACP gives tighter integration, with changes in Zed's diff review; the Terminal Thread keeps the plan's value.
Can several agents work on the same project at once?
Yes, as separate threads, each with its own agent, and Zed scopes them to a project and a git worktree. Put agents that edit files in separate worktrees: two agents changing the same checkout at once overwrite each other's work, while research or review threads can share one. The limit is usually review rather than the editor: three agents can produce more diffs in an afternoon than one person reads closely, which is where an AI code reviewer earns its place.
Stacks Related to Zed Dev Stack
Paseo Dev Stack
Developer
Run Claude Code, Codex, and other coding agents on your own machine with Paseo and steer them from any device.
Claude Code Dev Stack
DeveloperClaude Code in the terminal, desktop app, or web, paired with Claude and the add-ons to run it anywhere.
T3 Code Dev Stack
Developer
One open-source app for Claude Code, Codex, OpenCode, and other coding agents, on desktop and mobile.
Conductor Dev Stack
Developer
Run Claude Code, Codex, and Cursor's agent in parallel on your Mac with Conductor, each in its own workspace.
Scores
Popularity3/5
A well-known editor with an active open-source community and a strong following among performance-minded developers, though its user base trails Cursor and VS Code by a wide margin.
Learning Curve2/5
Opening a project and chatting with the agent is immediate, but Zed's own keymap and settings, model providers, external agents, and their separate sign-ins take some setting up.
Flexibility5/5
Six model families plus local models for its own agent, dozens of external agents over the Agent Client Protocol, any CLI in a Terminal Thread, MCP, and SSH remote development, all in an open-source editor.
Performance4/5
A native, GPU-rendered editor that stays fast on large codebases, while agent speed still tracks the chosen model and nothing runs once the laptop is off.
Portability5/5
Open source under GPL-3.0, plain files and git history, no lock-in to an agent or model, and the same agents run in other editors and terminals.
Tools in the Zed Dev Stack Stack
Zed Dev Stack Pricing
Zed itself is free, external agents included; Pro at $10/month adds unlimited edit predictions and $5 of hosted model usage, then list price plus 10%, and Business is $30 per seat. The Zed Agent can instead run on your own API keys or a ChatGPT plan. Each external agent bills through its own plan. GitHub and GitLab are free for individuals, local runtimes are free, and an optional server runs about €5.49–24/month.
Personal is free with 2,000 accepted edit predictions; Pro $10/mo with $5 of hosted usage, then list price +10%; Business $30/seat.
Instead of Zed-hosted models: e.g. DeepSeek V4.1-Flash $0.15/$0.60 per M off-peak, Grok 4.7 $2/$6, Claude Sonnet 5.5 $2/$10; a ChatGPT plan also works.
Claude Code from $20/mo (over ACP it uses Agent SDK credits at API rates); Codex and Copilot have free tiers, Copilot Pro $10/mo; Grok Build via SuperGrok $30/mo; OpenCode is free.
Free tier covers individual devs and small teams; GitHub Team is $4/mo, GitLab Premium is $29/mo for more seats.
Ollama, LM Studio, and llama.cpp are free; the cost is the hardware that runs the model.
Tailscale's personal tier is free.
4 GB covers Zed's remote server and language servers: about €5.49 at Hetzner, $6.49 intro at Hostinger ($11.99 renewal), $24 at DigitalOcean; or a Raspberry Pi as a one-time cost.