OmniFeed Search

Zed

Open-source native code editor with AI agents, edit predictions and broad provider support

Visit zed.dev
Screenshot of Zed
Zed, from Zed Industries

Our verdict

#4 of 7 in AI code editors · For fast, open AI coding

Pros

  • Excellent native editing with granular AI change controls
  • Broad choice of agents, providers, and local models
  • Free complete editor with opt-in training and strong privacy defaults

Cons

  • A much smaller extension ecosystem than VS Code
  • The native agent needs testing on your own repositories
  • Business identity and compliance features are still incomplete

Skip if: Teams that depend on specific VS Code extensions or require SSO, SCIM, and a current SOC 2 report.

Read the full review →

Rankings and alternatives

AI coding tools in 2027 scores

Evaluated Nov 12, 2026 Evidence ↓
  • Coding and repository-task quality · counts double 3

    3/5 means: Competitive on well-scoped tasks, but results or supporting evidence are less consistent than the leaders

    Native-agent users report successful issue implementation and one favorable same-model comparison with OpenCode, but there is no controlled multi-task comparison or broad practitioner-quality signal like the leaders. One contrary local-Ollama account also limits confidence. These records support competitive configured use rather than parity with the leading agents.

  • Everyday coding workflow 5

    5/5 means: Excellent completions and in-context review in a cohesive editor or across a broad choice of existing editors

    The fast native editor combines multi-line edit predictions, selected-code context, native and external agents, steering and granular hunk review in one interface. Its smaller extension ecosystem is the main tradeoff against VS Code-derived alternatives.

  • Agent control and recovery 4

    4/5 means: Strong approvals, isolation and automation, with a notable recovery or execution limitation

    Granular tool rules, individual-hunk review, steering, worktree isolation and parallel native or external agent threads provide a strong interactive control loop. It lacks the default OS sandbox and independent conversation/file recovery of the highest-scoring delegation workflows.

  • Model and provider choice · counts half 5

    5/5 means: Broad provider, subscription and custom-endpoint choice plus at least three named local-model runners

    Zed Agent supports hosted providers, API keys, subscriptions, gateways and custom endpoints, plus llama.cpp, LM Studio and Ollama for local inference. Claude subscriptions use an external agent rather than the native provider path.

  • Team administration and security 3

    3/5 means: Useful central policy and budget controls, but important identity, audit or networking features are limited

    Business can centrally restrict hosted models, edit predictions and data-sharing overrides, with direct-provider routes available. SSO, SAML, SCIM and SOC 2 are not currently available, leaving a material identity and assurance gap for managed rollouts.

  • Code and prompt privacy · counts half 5

    5/5 means: Direct-to-provider or local inference that the coding-tool vendor does not collect, with a documented no-training route

    Prompt sharing and edit-prediction training are opt-in, Zed says it does not store prompts, provider agreements prohibit training and direct API-key or local routes avoid Zed-hosted inference. A limited set of hosted provider models has short-term safety-retention exceptions.

AI code editors scores

Evaluated Nov 12, 2026 Evidence ↓
  • AI coding quality · counts double 3

    3/5 means: Useful everyday assistance with less consistency, maturity or supporting evidence than the leaders

    Native inline edits, configurable predictions and repository agents are useful, but the current outcome evidence is a small set of conflicting practitioner reports rather than a consistent record for complex work. External ACP agents do not raise this native-assistance score.

  • Editor quality 5

    5/5 means: An excellent development environment with outstanding speed or language-aware tools and no major compromise for its intended developers

    The outstanding native, GPU-rendered editor here for responsiveness and direct manipulation, with current debugging, remote development and collaboration tools. Its speed and simplicity are the clearest reasons to choose it over VS Code-derived editors.

  • Integrated AI workflow · counts double 4

    4/5 means: A cohesive manual-and-AI workflow with one significant rough edge or an opinionated interface that will not suit everyone

    Inline edits, predictions, native and external agents share a fast, coherent editor without imposing an agent board. The multiple agent and provider paths add setup choices, and its native AI polish does not have the same practitioner consensus as its core editing experience.

  • Review and change control 5

    5/5 means: Excellent granular review, redirection and recovery, with safe ways to isolate or verify larger changes

    Granular individual-hunk acceptance and rejection, editable diffs, tool permissions and worktree isolation provide excellent direct control. External-agent review can differ, but the native editor's controls are at the top of this comparison.

  • Compatibility with existing workflows 3

    3/5 means: Covers common setups, but important extensions, languages or remote workflows need checking

    Available on all major desktop platforms with modern debugging and remote tools, but its much smaller extension catalog leaves niche languages and enterprise toolchains less certain than VS Code or JetBrains.

  • Model and agent choice 5

    5/5 means: Multiple agent harnesses and providers, existing subscriptions or custom endpoints, and practical local inference

    Native and ACP agents, provider subscriptions, API keys, gateways and local models give the broadest practical choice here. Edit predictions also support alternatives to Zed's own open model.

  • Learning resources and support 3

    3/5 means: Adequate documentation and community help, with important AI or editor questions requiring specialist knowledge

    Current official documentation covers the editor, native AI, external agents and provider configuration. A smaller user and extension community leaves fewer established project-specific examples and troubleshooting answers than the long-standing editors.

  • Privacy and team controls 4

    4/5 means: Strong documented no-training choices and useful central controls, with a smaller gap in defaults, retention or administration

    Prompt sharing and training are opt-in, direct-provider routes avoid Zed's servers, and business admins can lock data policies. The absence of current SSO, SAML, SCIM and SOC 2 prevents a top score for managed enterprise adoption.

Pricing

  • Personal Free Full editor; use your own API keys or external agents
  • Pro $10.00 /month Unlimited hosted edit predictions and $5 of tokens; further usage billed at API list price plus 10%

From the vendor’s pricing page, checked Nov 12, 2026. That page is authoritative.

Evidence and sources

AI coding tools in 2027 scores

Coding and repository-task quality 3 quotes
  • “Personally I've found Zed's agent to be much superior to OpenCode. IDK why but it gets better results from the same model on the same task. Using with GLM 5.3 and DeepSeek 4.1.”
    denkmoon, Developer writing on Hacker News · news.ycombinator.com/item · checked
  • “Am I the only one using zed? Mostly I use it with Mistral but I also have deepseek in there. I find a good workflow is to store issues in markdown files and write thorough issue descriptions that reference relevant files. Then I can just ask the zed agent to implement the issue and it works pretty well.”
    tensor, Developer writing on Hacker News · news.ycombinator.com/item · checked
  • “Zed agent was so bad at using Ollama local models that I created my own agent to use instead. I think creating an agent is the hello-world of agentic coding, everyone is doing one.”
    Shorel, Developer writing on Hacker News · news.ycombinator.com/item · checked
Everyday coding workflow 6 quotes
  • “The default provider is Zeta, an open source model developed by Zed, but you can also use other providers like GitHub Copilot, Mercury Coder, and Codestral.”
    zed.dev/docs/ai/edit-prediction · checked
  • “Zed is the fastest editor on this list to open, type in, and render large files in, because it is written in Rust with GPU-accelerated rendering rather than wrapping a browser engine.”
    guptadeepak.com/tools/top-5-code-editors-ides-2026 · checked
  • “The extension marketplace sits around 800 extensions versus VS Code's 50,000-plus, so niche languages and enterprise toolchains have real coverage gaps.”
    guptadeepak.com/tools/top-5-code-editors-ides-2026 · checked
  • “You can accept or reject each individual change hunk, or the whole set of changes made by the agent.”
    zed.dev/docs/ai/agent-panel · checked
  • “Edit Prediction is how Zed's AI code completions work: an LLM predicts the code you want to write. As you type, Zed requests predictions from the edit prediction provider, which returns individual or multi-line suggestions you accept by pressing tab.”
    zed.dev/docs/ai/edit-prediction · checked
  • “You can run multiple agent threads at once, each working independently with its own agent, context window, and conversation history.”
    zed.dev/docs/ai/agent-panel · checked
Agent control and recovery 3 quotes
  • “You can accept or reject each individual change hunk, or the whole set of changes made by the agent.”
    zed.dev/docs/ai/agent-panel · checked
  • “Auto-deny dangerous actions (blocked even when tool_permissions.default is set to "allow")”
    zed.dev/docs/ai/tool-permissions · checked
  • “You can run multiple agent threads at once, each working independently with its own agent, context window, and conversation history.”
    zed.dev/docs/ai/agent-panel · checked
Model and provider choice 3 quotes
  • “Zed Agent: Zed’s native agent. It can use models configured through LLM Providers, including Zed-hosted models, provider API keys, supported subscriptions, gateways, and local models. It also uses built-in tools, profiles, skills, instructions, and MCP servers. External Agents: ACP-integrated agents that run through their own process and configuration.”
    zed.dev/docs/ai/overview · checked
  • “| llama.cpp | Yes | Separate config | Separate config | Configure a llama.cpp server for Zed AI features | | LM Studio | Yes | Separate config | Separate config | Configure LM Studio for Zed AI features | | Ollama | Yes | Separate config | Separate config | Configure Ollama for Zed AI features |”
    zed.dev/docs/ai/use-a-local-model · checked
  • “| ChatGPT Plus / Pro | ChatGPT Subscription | Codex where supported | Codex CLI | Sign in with OpenAI in Zed; separate from OpenAI API keys | | Claude Pro / Max | No direct Zed LLM provider path | Claude Agent | Claude Code | Separate from Anthropic API keys | | GitHub Copilot | GitHub Copilot Chat; Copilot edit prediction | Copilot agent where available | CLI where available | Requires Copilot/Copilot Chat auth |”
    zed.dev/docs/ai/use-an-existing-subscription · checked
Team administration and security 2 quotes
  • “SSO, SAML, SCIM, and SOC 2 are not currently available, but we're working on them now.”
    zed.dev/business · checked
  • “Business admins can restrict or disable Zed-hosted AI models by model, turn off edit predictions org-wide, and lock data sharing settings so members can't override them.”
    zed.dev/pricing · checked
Code and prompt privacy 2 quotes
  • “Zed doesn't store prompts or use them to train models, and our agreements with AI providers prohibit training on your data. Zero data retention agreements cover all hosted models, except a limited set of provider-designated models, such as Anthropic's Mythos-class models, where the provider retains prompts and outputs for a limited period for safety purposes (details). Prompt sharing and edit prediction training are both opt-in for every Zed user.”
    zed.dev/business · checked
  • “| llama.cpp | Yes | Separate config | Separate config | Configure a llama.cpp server for Zed AI features | | LM Studio | Yes | Separate config | Separate config | Configure LM Studio for Zed AI features | | Ollama | Yes | Separate config | Separate config | Configure Ollama for Zed AI features |”
    zed.dev/docs/ai/use-a-local-model · checked

AI code editors scores

AI coding quality 5 quotes
  • “Zed Agent: Zed’s native agent. It can use models configured through LLM Providers, including Zed-hosted models, provider API keys, supported subscriptions, gateways, and local models. It also uses built-in tools, profiles, skills, instructions, and MCP servers. External Agents: ACP-integrated agents that run through their own process and configuration.”
    zed.dev/docs/ai/overview · checked
  • “The default provider is Zeta, an open source model developed by Zed, but you can also use other providers like GitHub Copilot, Mercury Coder, and Codestral.”
    zed.dev/docs/ai/edit-prediction · checked
  • “Personally I've found Zed's agent to be much superior to OpenCode. IDK why but it gets better results from the same model on the same task. Using with GLM 5.3 and DeepSeek 4.1.”
    denkmoon, Developer writing on Hacker News · news.ycombinator.com/item · checked
  • “Am I the only one using zed? Mostly I use it with Mistral but I also have deepseek in there. I find a good workflow is to store issues in markdown files and write thorough issue descriptions that reference relevant files. Then I can just ask the zed agent to implement the issue and it works pretty well.”
    tensor, Developer writing on Hacker News · news.ycombinator.com/item · checked
  • “Zed agent was so bad at using Ollama local models that I created my own agent to use instead. I think creating an agent is the hello-world of agentic coding, everyone is doing one.”
    Shorel, Developer writing on Hacker News · news.ycombinator.com/item · checked
Editor quality 1 quote
  • “Zed is the fastest editor on this list to open, type in, and render large files in, because it is written in Rust with GPU-accelerated rendering rather than wrapping a browser engine.”
    guptadeepak.com/tools/top-5-code-editors-ides-2026 · checked
Integrated AI workflow 3 quotes
  • “Zed Agent: Zed’s native agent. It can use models configured through LLM Providers, including Zed-hosted models, provider API keys, supported subscriptions, gateways, and local models. It also uses built-in tools, profiles, skills, instructions, and MCP servers. External Agents: ACP-integrated agents that run through their own process and configuration.”
    zed.dev/docs/ai/overview · checked
  • “The default provider is Zeta, an open source model developed by Zed, but you can also use other providers like GitHub Copilot, Mercury Coder, and Codestral.”
    zed.dev/docs/ai/edit-prediction · checked
  • “You can accept or reject each individual change hunk, or the whole set of changes made by the agent.”
    zed.dev/docs/ai/agent-panel · checked
Review and change control 1 quote
  • “You can accept or reject each individual change hunk, or the whole set of changes made by the agent.”
    zed.dev/docs/ai/agent-panel · checked
Compatibility with existing workflows 2 quotes
  • “macOS Apple SiliconLoading… Requires version 12.0 or later Windows Intel/AMDLoading… Linux curl -f https://zed.dev/install.sh | sh”
    zed.dev/download · checked
  • “The extension marketplace sits around 800 extensions versus VS Code's 50,000-plus, so niche languages and enterprise toolchains have real coverage gaps.”
    guptadeepak.com/tools/top-5-code-editors-ides-2026 · checked
Model and agent choice 4 quotes
  • “Personal $0forever The next-generation code editor. Free forever. Unlimited use with your API keys or external agents like Claude Agent, Codex CLI, and more”
    zed.dev/pricing · checked
  • “Zed Agent: Zed’s native agent. It can use models configured through LLM Providers, including Zed-hosted models, provider API keys, supported subscriptions, gateways, and local models. It also uses built-in tools, profiles, skills, instructions, and MCP servers. External Agents: ACP-integrated agents that run through their own process and configuration.”
    zed.dev/docs/ai/overview · checked
  • “The default provider is Zeta, an open source model developed by Zed, but you can also use other providers like GitHub Copilot, Mercury Coder, and Codestral.”
    zed.dev/docs/ai/edit-prediction · checked
  • “Zeta predictions are included with Zed Pro, the Pro trial, and the Student and Business plans. On the Free plan, you can use other providers, like GitHub Copilot or a local model.”
    zed.dev/docs/ai/edit-prediction · checked
Learning resources and support 6 quotes
  • “Zed Agent: Zed’s native agent. It can use models configured through LLM Providers, including Zed-hosted models, provider API keys, supported subscriptions, gateways, and local models. It also uses built-in tools, profiles, skills, instructions, and MCP servers. External Agents: ACP-integrated agents that run through their own process and configuration.”
    zed.dev/docs/ai/overview · checked
  • “The extension marketplace sits around 800 extensions versus VS Code's 50,000-plus, so niche languages and enterprise toolchains have real coverage gaps.”
    guptadeepak.com/tools/top-5-code-editors-ides-2026 · checked
  • “Zed | 1,504 | 11.4%”
    survey.stackoverflow.co/2026/technology/data/dev-ide · checked
  • “Auto-deny dangerous actions (blocked even when tool_permissions.default is set to "allow")”
    zed.dev/docs/ai/tool-permissions · checked
  • “| llama.cpp | Yes | Separate config | Separate config | Configure a llama.cpp server for Zed AI features | | LM Studio | Yes | Separate config | Separate config | Configure LM Studio for Zed AI features | | Ollama | Yes | Separate config | Separate config | Configure Ollama for Zed AI features |”
    zed.dev/docs/ai/use-a-local-model · checked
  • “| ChatGPT Plus / Pro | ChatGPT Subscription | Codex where supported | Codex CLI | Sign in with OpenAI in Zed; separate from OpenAI API keys | | Claude Pro / Max | No direct Zed LLM provider path | Claude Agent | Claude Code | Separate from Anthropic API keys | | GitHub Copilot | GitHub Copilot Chat; Copilot edit prediction | Copilot agent where available | CLI where available | Requires Copilot/Copilot Chat auth |”
    zed.dev/docs/ai/use-an-existing-subscription · checked
Privacy and team controls 3 quotes
  • “SSO, SAML, SCIM, and SOC 2 are not currently available, but we're working on them now.”
    zed.dev/business · checked
  • “Zed doesn't store prompts or use them to train models, and our agreements with AI providers prohibit training on your data. Zero data retention agreements cover all hosted models, except a limited set of provider-designated models, such as Anthropic's Mythos-class models, where the provider retains prompts and outputs for a limited period for safety purposes (details). Prompt sharing and edit prediction training are both opt-in for every Zed user.”
    zed.dev/business · checked
  • “Business admins can restrict or disable Zed-hosted AI models by model, turn off edit predictions org-wide, and lock data sharing settings so members can't override them.”
    zed.dev/pricing · checked

Facts

License 1 quote
Runs on 1 quote
  • “macOS Apple SiliconLoading… Requires version 12.0 or later Windows Intel/AMDLoading… Linux curl -f https://zed.dev/install.sh | sh”
    zed.dev/download · checked
Needs 1 quote
  • “Personal $0forever The next-generation code editor. Free forever. Unlimited use with your API keys or external agents like Claude Agent, Codex CLI, and more”
    zed.dev/pricing · checked

Other quotes

Not cited by a score or fact 3 quotes
  • “Pro Free Trial $10per month The next-generation code editor + AI. Supercharge your development. Unlimited edit predictions $5 of tokens included Usage-based billing beyond $5”
    zed.dev/pricing · checked
  • “A token is a unit of text processed by an LLM (roughly 3–4 characters of English text). With token-based billing, you’re only charged for the text your prompts and responses actually consume. Costs vary by model, but Zed Pro bills at API list price +10% for hosted usage.”
    zed.dev/pricing · checked
  • “In my setup I use multiple agents like Claude Code and Codex, and Zed’s ACP support makes it pretty nice to manage them all as “threads” in one place. Worktree switching also feels much smoother.”
    turastory, Developer writing on Hacker News · news.ycombinator.com/item · checked

Pricing

zed.dev/pricing · checked