DevPrice
中文
Back to previous page
CCursor logo

Cursor

An AI-native code editor with completion, codebase chat, agents, and edits.

AI codingCode assistantsCoding agentsFree plan
Official website

Plans and pricing

Official pricing verified
Access
Free plan
Paid plans from
$20/mo
Billing
Monthly subscription

Hobby

$0

Pro

$20/mo

Pro+

$60/mo

Ultra

$200/mo

Official pricing

Official sources

Cursor official websiteCursor PricingCursor DocumentationCursor Agent OverviewCursor VS Code MigrationCursor Agents WindowCursor Plan ModeCustomize Cursor

Verified: September 6, 2026

DevPrice organizes public information and does not sell Cursor subscriptions. Prices and availability are determined by Cursor.

What is Cursor?

Cursor has moved beyond being a code editor with an AI chat panel. It is an AI development environment organized around real repositories. Built on the VS Code codebase, it can retain familiar extensions, themes, settings, and keybindings, while the agent-first Agents Window gives developers a higher-level workspace for managing complete software tasks without giving up a conventional editor.

For focused work, Cursor Tab predicts what should change after recent edits, while Inline Edit applies natural-language instructions to selected code. When a request expands across files or systems, Plan Mode can first inspect the repository, clarify requirements, and produce an editable implementation plan. Agent can then search the codebase, edit files, run commands, and react to the resulting checks. Cursor also treats agents as ongoing units of work rather than one-off chat replies. Multiple agents can operate in parallel in isolated environments, and local work can be handed to a Cloud Agent to continue remotely. Rules, Skills, Hooks, plugins, and MCP connections provide project conventions, repeatable procedures, and external tools so a team does not need to explain its working style from scratch in every task. This workflow is best suited to developers and teams that repeatedly need to understand, change, and validate an existing codebase: onboarding to an unfamiliar project, implementing a feature across modules, investigating an elusive defect, or carrying out a large refactor. Cursor can absorb much of the context gathering and repetitive execution, but its output can still miss the intended result; diffs, tests, review, and permanent version control remain essential before shipping.

Key features of Cursor

  • Switch between the editor and Agents Window

    The classic editor is suited to direct file work, while Agents Window brings together agent tasks across repositories and local, cloud, or remote environments. Developers can switch between the two views or keep both open.

  • From next-edit prediction to focused changes

    Tab predicts follow-up edits from the current code and recent changes for fast everyday coding. Inline Edit handles clearly bounded work by applying generation, modification, or refactoring instructions to selected code.

  • Understand the repository before changing it

    Agent can search file names, directory structures, and exact code patterns, then read the implementation relevant to a task. That context helps explain unfamiliar modules, trace call paths, and locate the files a change actually affects.

  • Create a reviewable plan for complex work

    Plan Mode clarifies the request and studies the repository before producing an editable implementation plan. It is useful for cross-file features, architectural changes, or uncertain scope that should be reviewed before code is written.

  • Carry a request through to reviewable changes

    Agent can read and edit files, execute terminal commands, search the web, and continue correcting its work from build, test, or runtime output instead of returning a snippet to copy. The resulting changes remain available as diffs for developer approval.

  • Run local and cloud tasks in parallel

    Agents Window can run multiple isolated tasks at once, reducing interference between their changes. Long-running work can continue in a remote Cloud Agent and return as a diff or pull request for review.

  • Turn team conventions into reusable workflows

    Rules and AGENTS.md preserve coding conventions and project context; Skills, Commands, and Hooks encode repeated procedures; plugins and MCP connect external tools and data. These settings can be managed for a user, workspace, or team.

  • Review, verify, and recover when work drifts

    Developers can inspect Agent changes in a diff view and use builds, tests, and code review to decide whether to accept them. Automatic checkpoints can undo a wrong turn within an agent session, but they restore files only and should not replace Git.