52 facts per product, each taken from the vendor's own public pages and linked to its source. 23 rows differ and are highlighted; the better documented value in a row is bold. “Not documented” means the vendor's pages do not state it — it is never counted as a “No”.
Facts verified · Prices in USD · Machine-readable JSON
Zed Pro is $10 per month, with unlimited edit predictions and $5 of hosted-model tokens included, then usage at API list price plus 10%. Cursor Pro is $16 per month billed yearly or $20 monthly, with a monthly usage allowance. Both have a free plan.
Zed accepts your own API keys on every plan, including free, and runs local models through llama.cpp, LM Studio, Ollama or any local OpenAI-compatible server. Cursor accepts OpenAI, Anthropic, Google, Azure OpenAI and AWS Bedrock keys for chat models only, with requests passing through Cursor's backend; its pages do not document local models.
Zed is open source under GPL-3.0-or-later, with Apache-2.0 components, and is its own editor written in Rust. Cursor is a proprietary editor.
Cursor's Cloud Agents run in isolated cloud VMs, in parallel, started from the desktop, the web, the iOS app, Slack, GitHub or Linear. Zed's Parallel Agents run as local threads inside the editor, and its pages document no cloud agents, web app or mobile app.
Cursor's Bugbot reviews pull requests automatically or on demand, billed from included or on-demand usage. Zed's pages do not document pull request review.
Cursor documents SOC 2 Type II, ISO 27001:2022, HIPAA BAAs on Enterprise, and SSO on Teams. Zed states SSO, SAML and SCIM are not available yet, it is working toward SOC 2 Type 1, and its Terms prohibit HIPAA protected health information.
| Cursor | Zed | |
|---|---|---|
| Free plan | Yes Hobby plan; no credit card required. | Yes Personal plan, free forever. AI features on this plan run on your own API keys, existing subscriptions, local models or external agents, so model usage is paid to those providers. |
| Free plan limits | Limited Agent requests; access to Composer; monthly Tab allowance Help center: Hobby covers Agent, Chat and Tab completions with the Auto model and a smaller model set; Tab suggestions pause once the monthly allowance is used; free accounts cannot start Cloud Agents. | 2,000 accepted edit predictions; no Zed-hosted AI models (own API keys or external agents only) The Edit Prediction docs state the free plan includes 2,000 Zeta predictions per month. A Student plan (unlimited edit predictions, $10/month token credit) is free for one year to verified university students; it is listed in the docs, not on the pricing page. |
| Free trial (days) | Not documented Not documented on vendor pages | 14 Pro trial. Ends after 14 days or when the $5 trial balance is used, whichever comes first; one hosted model only (the pricing page names GPT-5.6 Luna, the docs name GPT-6 Luna). The trial balance is shared with Delta. Converts to the free plan when it ends. Business has no free trial. |
| Trial requires credit card | Not documented Not documented on vendor pages | No Pro trial. |
| Cheapest paid plan, billed annually (per user / month) | $16.00 Pro (individual plan), billed yearly. An India-only Start plan (₹649/mo, billed monthly in INR) is listed in the help center, not on the pricing page. | Not documented Not documented on vendor pages. Plans are billed monthly and no annual price is listed. |
| Cheapest paid plan, billed monthly (per user / month) | $20.00 Pro (individual plan), billed monthly. An India-only Start plan (₹649/mo, INR) is listed in the help center, not on the pricing page. | $10.00 Pro (individual plan), billed monthly. Includes $5 of tokens per month; further usage billed at API list price +10%. |
| Highest listed-price plan, billed annually (per user / month) | $160.00 Ultra (individual plan), billed yearly. Highest per-seat team price is Teams Premium at $96/user/mo billed yearly. | Not documented Not documented on vendor pages. No annual prices are listed; the highest listed price is Business at $30 per seat per month. |
| Minimum paid seats | Not documented Not documented on vendor pages | 1 No seat minimum for Business; it is self-serve for organizations of any size. Order-form contracts start at 25 seats. |
| Custom-priced enterprise plan | Yes Enterprise is contact-sales; priced per seat with an included usage allotment and optional pre-committed usage. | Yes No separate Enterprise tier. The Business page invites organizations to email about custom Business plans; contracts of 25+ seats with invoice billing go through sales@zed.dev. |
| Cursor | Zed | |
|---|---|---|
| Web app | Yes Browser access at cursor.com/agents for starting, following and reviewing cloud agents. | Not documented Not documented on vendor pages |
| Runs on Windows | Yes Desktop app for Windows x64 and ARM64 (user or system installer); CLI via WSL or native PowerShell. | Yes Windows 10 (1903+) and Windows 11 (22H2+); x64 and Arm64. |
| Runs on macOS | Yes Desktop app for Apple silicon, Intel and Universal builds; CLI via install script. | Yes Intel and Apple Silicon; the download page states macOS 10.15 or later. |
| Runs on Linux | Yes Desktop app as .deb, RPM and AppImage for x64 and ARM64; CLI via install script. | Yes x86_64 and aarch64; install script or distribution packages. |
| iOS app | Yes Cursor for iOS on iPhone and iPad (iOS/iPadOS 26.0+) directs and reviews cloud agents and Remote Control sessions; no editor or terminal. Starting agents needs a paid plan. | Not documented Not documented on vendor pages |
| Android app | No No Cursor app for Android; an Android version is planned with no release date. The separate Grok Bot app (a general-purpose agent included with paid Cursor plans) is on Google Play. | Not documented Not documented on vendor pages |
| Browser extension | Not documented Not documented on vendor pages | Not documented Not documented on vendor pages |
| Cursor | Zed | |
|---|---|---|
| Own standalone code editor | Yes Desktop app with an agent-first Agents Window plus a classic editor that runs VS Code extensions from the Open VSX registry and imports VS Code settings, keybindings and extensions. The vendor pages do not call it a fork. | Yes Zed is its own editor, written from scratch in Rust (not a fork of another editor). |
| Extension for VS Code | Not documented Not documented on vendor pages | Not documented Not documented on vendor pages |
| Plugin for JetBrains IDEs | Yes Through the Agent Client Protocol: install Cursor from the ACP registry in the JetBrains AI Chat panel (AI Assistant plugin, 2025.1+). Paid Cursor plan required. Provides the agent (model choice, file edits, terminal commands); Tab autocomplete is not listed. | Not documented Not documented on vendor pages |
| Command-line (terminal) agent | Yes Cursor CLI (`agent`): interactive, Plan and Ask modes plus headless mode for scripts, CI and GitHub Actions. Installs on macOS, Linux, Windows (WSL and native PowerShell). | Not documented Not documented on vendor pages. The zed command opens files, folders and diffs but is not an agent; Terminal Threads run third-party agent CLIs inside Zed. |
| Other editors supported | JetBrains IDEs (via ACP); any editor or IDE through the Cursor CLI in a terminal; custom ACP clients The Xcode integration is an MCP bridge that lets Cursor's agent build, test and edit Xcode projects from Cursor; it is not a plugin inside Xcode. | Not documented Not documented on vendor pages |
| Cursor | Zed | |
|---|---|---|
| Inline code completion (autocomplete) | Yes Tab. Unlimited on Pro, Pro+ and Ultra; Hobby has a monthly Tab allowance. Tab always uses Cursor's own models, even with your own API keys. | Yes Edit Prediction. Default provider is Zeta, Zed's open model; GitHub Copilot, Mercury Coder, Codestral and local models can be used instead. Free plan: 2,000 predictions per month; unlimited on Pro and Business. |
| Agent that edits across multiple files | Yes Agent, Plan, Ask and Debug modes in the editor, Agents Window, CLI, web and iOS. | Yes Zed Agent (native), plus External Agents over the Agent Client Protocol (Claude, Codex, Copilot, Cursor, OpenCode and others). |
| Agent runs terminal commands | Yes Run Mode sets when commands run automatically, ask for approval, or run in a sandbox. | Yes Built-in terminal tool, governed by tool permissions; optional OS-level sandbox. |
| Cloud / background agents | Yes Cloud Agents run in isolated cloud VMs, in parallel, from desktop, web, iOS, Slack, GitHub and Linear; Automations run agents on schedules or triggers. Included on paid plans (Pro and up); free accounts cannot start them. Tool execution can run on Self-Hosted Machines. | Not documented Not documented on vendor pages. Parallel Agents run several local agent threads at once inside the editor; no cloud or remote agent runner is documented for Zed. |
| Pull request code review | Yes Bugbot reviews PRs automatically or on demand. Usage-based billing: individual plans draw from included usage, then on-demand; Teams bill from on-demand spend. Not included in the India-only Start plan. | Not documented Not documented on vendor pages |
| Codebase indexing / semantic code search | Yes Local Instant Grep index for exact and regex code search, plus an Explore subagent for broad searches. Cursor states it does not upload code or store codebase embeddings for search. | Not documented Not documented on vendor pages. The agent's documented search tools are grep and path-glob search. |
| MCP (Model Context Protocol) servers | Yes Pricing page lists MCPs, skills and hooks among Pro features; Enterprise admins can allowlist or block MCP servers. | Yes MCP servers for the Zed Agent; Zed can forward them to External Agents over ACP. MCP servers can also be installed as extensions. |
| Project rules / custom instruction files | Yes Project rules in .cursor/rules (.mdc files), AGENTS.md, user rules, and enforceable team rules on Teams and Enterprise. | Yes Personal and project AGENTS.md; Zed also reads .rules, .cursorrules, .windsurfrules, .clinerules, .github/copilot-instructions.md, AGENT.md, CLAUDE.md and GEMINI.md. Skills are supported too. |
| Checkpoints to roll back agent edits | Yes Created automatically before significant Agent changes; stored locally, separate from Git; restoring reverts files only. | Yes Restore Checkpoint on each agent edit. Availability for External Agents depends on the agent integration. |
| Cursor | Zed | |
|---|---|---|
| User can choose the AI model | Yes Pick a model or use Auto (Cursor Router). Hobby gets a smaller model set; paid plans unlock all models; Enterprise admins can restrict models. | Yes Model selector in the Agent Panel across Zed-hosted, API-key, subscription, gateway and local providers. |
| Model families offered | Cursor models (Grok 4.7, 4.6, 4.5; Composer 2.5), OpenAI GPT, Anthropic Claude, Google Gemini, and others Grok models are listed as Cursor models in the Cursor Models usage pool. The docs model table also lists Meta, Moonshot and Z.ai models. | Anthropic Claude, OpenAI GPT and Google Gemini hosted by Zed; other providers via own keys, gateways or local models Hosted list includes Claude Fable, Opus, Sonnet and Haiku; GPT-6, GPT-5.x and GPT-5.x-Codex; Gemini 3.x. Hosted xAI models were retired on May 15, 2026. Own-key providers include Amazon Bedrock, DeepSeek, Mistral, xAI, OpenRouter, Vercel AI Gateway and GitHub Copilot. |
| Bring your own API key | Yes OpenAI, Anthropic, Google, Azure OpenAI and AWS Bedrock keys; chat models only (Tab stays on Cursor models). Requests still pass through Cursor's backend. Individual plans: not counted against included usage. Teams/Enterprise: $0.25 per million tokens Cursor Token Rate still applies. Enterprise admins can disable BYOK. Cursor's zero data retention does not cover BYOK requests. | Yes All plans, including free. Keys are stored in the system keychain. Business admin controls do not cover own-key use. |
| Local / self-hosted models | Not documented Not documented on vendor pages | Yes llama.cpp, LM Studio, Ollama and local OpenAI-compatible servers; local edit prediction is also supported. |
| How paid-plan usage is limited | Monthly included model usage in two pools (Cursor Models; Other Models at API prices), then optional on-demand pay-as-you-go at API rates Pro+ has 3x and Ultra 20x Pro Agent limits; Teams Premium has 5x Standard. Usage resets monthly without rollover. Teams and Enterprise add a $0.25 per million tokens Cursor Token Rate on third-party models. Enterprise offers pooled usage. | Pro: $5/month token credit, then billed at API list price +10% up to a monthly spend limit (default $10 extra) Hosted usage is metered in tokens. Pro users can set the extra-spend limit to $0 or raise it. Business has no bundled credits and an admin-set org-wide spend limit that starts at $0. Student plan usage is capped at its $10/month credit. |
| Open-source client | No Proprietary client. The public github.com/cursor/cursor repository holds a README and issue templates, not the app source. Some auxiliary repos are open source (e.g. plugins, SDK bridge, Terraform provider). | Yes GPL-3.0-or-later, with Apache-2.0 components. |
| Cursor | Zed | |
|---|---|---|
| Zero data retention option | Yes Applies with Privacy Mode on, available on every plan including free; on by default for Enterprise, enforceable org-wide by Teams and Enterprise admins. Exceptions: your own API keys, models that require provider retention (admin approval needed), and prompts flagged by providers' abuse classifiers. | Yes Default on all plans for Zed-hosted models, except provider-designated models: Anthropic retains Claude Fable 5.1 and Claude Fable 5 prompts and outputs for at least 30 days for safety. Zed keeps AI data only when a user opts in to feedback or training-data sharing; Business turns both off by default. Own keys, gateways and external agents follow each provider's terms. |
| Code excluded from model training | Yes Requires Privacy Mode. With Privacy Mode off, Cursor may use code data to train its models. Terms: Anysphere does not train on your content unless you explicitly agree. | Yes Provider agreements for Zed-hosted models prohibit training. Edit Prediction training data is collected only if a user opts in and the project is open source; on Business, opt-in is off by default and admin-controlled. |
| Self-hosted or on-premises deployment | No No on-premises deployment; the service runs on AWS. Cloud Agent tool execution can move to Self-Hosted Machines you manage, while Cursor still runs the agent loop and inference. AWS PrivateLink and Cloudflare Tunnel private connectivity are available. | Not documented Not documented on vendor pages |
| Team admin dashboard and usage analytics | Yes Teams plan and up: centralized billing and administration with usage analytics. Enterprise adds SCIM, audit logs, AI Code Tracking API and Conversation Insights. | Yes Business plan. dashboard.zed.dev for members and roles, consolidated billing, org-wide AI spend limit, per-member hosted AI spend tracking, and toggles for hosted models, edit predictions, feedback sharing and collaboration. |
| Cursor | Zed | |
|---|---|---|
| Listed integrations | 370 Plugins listed in the Cursor Marketplace (MCP servers, skills and tool bundles from third parties and Cursor), counted as unique plugin listings; excludes 27 automation templates. Separately, the docs cover 11 native integrations: GitHub, GitLab, Azure DevOps, Bitbucket, Slack, Microsoft Teams, Jira, Linear, Notion, JetBrains and Xcode. | 73 MCP server extensions listed under the MCP Servers filter of the Zed extensions gallery. The gallery states no total and also lists languages, themes, debug adapters and agent servers ("hundreds" of extensions overall). |
| Public API | Yes Cloud Agents API (beta, all plans), TypeScript and Python SDKs (all users); Admin, Analytics, AI Code Tracking and Bugbot APIs for Enterprise teams. | Not documented Not documented on vendor pages. The documented extension API and Agent Client Protocol are for building extensions and agents, not for driving Zed programmatically. |
| Single sign-on (SSO) | Yes SAML/OIDC SSO on Teams and Enterprise at no extra cost; SCIM provisioning on Enterprise. | No SSO, SAML and SCIM are planned but not available. |
| Two-factor authentication | Not documented Not documented on vendor pages | Not documented Not documented on vendor pages. Sign-in uses GitHub OAuth. |
| SOC 2 report | Yes SOC 2 Type II; report available on request via trust.cursor.com. | No Not available yet; the docs say Zed is working toward SOC 2 Type 1. |
| ISO 27001 certification | Yes ISO/IEC 27001:2022; also ISO/IEC 42001:2023 and AIUC-1. | Not documented Not documented on vendor pages |
| HIPAA compliance offered | Yes Enterprise only: signed BAA required, with Privacy Mode enabled and locked org-wide. The standard Terms of Service prohibit sending HIPAA-regulated data otherwise. | No The Terms prohibit putting HIPAA protected health information into the service. |
| Cursor | Zed | |
|---|---|---|
| Support channels | Community forum, email (hi@cursor.com), help center with a signed-in support request form; priority support on Enterprise Individual and Teams plans: community and standard support. Enterprise: first human response within 8 hours (critical) or 24 hours (standard). | Email, GitHub Issues, Discord, GitHub Discussions, Reddit Email: hi@zed.dev (general), billing-support@zed.dev (billing), sales@zed.dev, security@zed.dev. Editor bugs go to GitHub Issues. Business support requests are prioritized. |
| 24/7 support | Not documented Not documented on vendor pages | Not documented Not documented on vendor pages |
| Vendor (legal entity) | Anysphere, Inc. Cursor's blog announced on Aug 14, 2026 that Cursor has been acquired by SpaceX; the Terms of Service still name Anysphere, Inc. | Zed Industries, Inc. Terms governed by Delaware law. |
| Founded | Not documented Not documented on vendor pages | Not documented Not documented on vendor pages |
| Headquarters | San Francisco, CA, USA Correspondence address in the Terms of Service: 2261 Market Street STE 86466, San Francisco, CA 94114. The vendor pages do not name a headquarters. | Denver, CO, USA Correspondence address in the Terms of Service: 2590 Welton Street, Suite 200, PMB 1916, Denver, CO 80205. The vendor pages do not name a headquarters, and the jobs page describes a fully remote team. |
| Cursor plan | Billed annually | Billed monthly |
|---|---|---|
| Hobby Per month Free individual plan. No credit card required; limited Agent requests; access to Composer. | Free | Free |
| Pro Per month Individual plan. $16/mo billed yearly, $20/mo billed monthly. Extended limits on Agent, access to frontier models, MCPs, skills and hooks, Cloud agents, Grok Bot access; Bugbot on usage-based billing. Includes monthly usage in two pools (Cursor Models; Other Models at API prices), then optional on-demand usage at API rates. | $16.00 | $20.00 |
| Pro+ Per month Individual plan. $48/mo billed yearly, $60/mo billed monthly. Everything in Pro, plus 3x Pro limits on Agent and higher Grok Bot usage limits. | $48.00 | $60.00 |
| Ultra Per month Individual plan. $160/mo billed yearly, $200/mo billed monthly. Everything in Pro, plus 20x Pro limits on Agent, highest Grok Bot usage limits and priority access to new features. | $160.00 | $200.00 |
| Teams (Standard seat) $32/user/mo billed yearly, $40/user/mo billed monthly. Centralized team billing and administration, team marketplace for rules, skills and plugins, cloud agents and automations with shared team context, agentic code reviews with Bugbot, usage analytics, team-wide privacy mode, SAML/OIDC SSO. Per-seat included usage; on-demand usage beyond it. | $32.00 | $40.00 |
| Teams (Premium seat) $96/user/mo billed yearly, $120/user/mo billed monthly. Same Teams features with 5x Standard limits on Agent and higher Grok Bot usage limits. Standard and Premium are seat types within the Teams plan. | $96.00 | $120.00 |
| Enterprise Custom (contact sales) Everything on Teams plus pooled usage, invoice/PO billing, SCIM seat management, repository/model/MCP access controls, auto-run, browser and network controls, audit logs and service accounts, AI code tracking API, priority support and account management. | Custom | Custom |
| Zed plan | Billed annually | Billed monthly |
|---|---|---|
| Personal Forever Free editor. 2,000 accepted edit predictions; unlimited AI use with your own API keys or external agents (Claude Agent, Codex CLI and others). No Zed-hosted models. | Free | Free |
| Pro Per month Individual plan, billed monthly; no annual price is listed. Unlimited edit predictions, $5 of tokens included each month, usage-based billing beyond $5 at API list price +10%. 14-day free trial, no credit card. | Not offered | $10.00 |
| Business Per seat, per month Month-to-month; no annual price is listed. No bundled AI credits: Zed-hosted AI billed at standard usage rates, or bring your own keys. Org-wide AI model policies, data governance controls, unified spend visibility, unlimited edit predictions, role-based access controls. No seat minimum; order forms available at 25+ seats. No free trial. | Not offered | $30.00 |
Per user per month unless the plan states otherwise, as listed on each vendor's pricing page.