# Cursor vs Claude Code: pricing, models, and cloud agents compared

Compare Cursor and Claude Code plans side by side, what Claude models cost in each, where each one runs, and how Cursor Cloud Agents and Claude Code cloud sessions differ.

- Published: 2026-11-09
- Updated: 2026-11-10
- Canonical: https://replicas.dev/resources/cursor-vs-claude-code

Cursor is an AI code editor made by Anysphere. It also runs agents from a CLI and in the cloud, and it can use models from several vendors. Claude Code is Anthropic's coding agent. It runs in the terminal, in code editors, in a desktop app and in the browser, always with Claude models.

The two products overlap more every month, but they still charge for different things. This guide compares where each one runs, what the plans include, and how their cloud agents work. Every claim links to the vendor's own page, and we checked them all on November 9, 2026.

## Which one to pick

Most teams decide on two questions: where they want to work, and which models they want to pay for.

| If you want | Pick | Why |
|---|---|---|
| An AI code editor | Cursor | Cursor is a full editor with its agent built in |
| A terminal agent that works with any editor | Claude Code | It runs in the terminal and has extensions for VS Code, Cursor and JetBrains |
| Claude models at a flat monthly price | Claude Code | Claude plans include Claude model usage. Cursor bills Claude models on top of the plan |
| Models from several vendors in one tool | Cursor | Composer, Grok, Claude, GPT and Gemini are all available |
| The lowest entry price | Cursor | Plus costs $8 a month, but covers only Cursor's own models |

You can also use both. Anthropic's Claude Code extension installs in Cursor as well as VS Code ([Anthropic](https://code.claude.com/docs/en/overview)), so you can keep Cursor's editor and run Claude Code on a Claude plan in the same window.

## Where each one runs

| | Cursor | Claude Code |
|---|---|---|
| Main surface | A desktop code editor | A terminal CLI |
| Also runs in | The `agent` CLI, the web, the iOS app, Slack, Linear, GitHub and Bitbucket comments, and an API ([Cursor](https://cursor.com/docs/cloud-agent)) | VS Code, JetBrains, a desktop app, the web at claude.ai/code, the Claude mobile app and Slack ([Anthropic](https://code.claude.com/docs/en/overview)) |
| In CI | The CLI's print mode ([Cursor](https://cursor.com/docs/cli/overview)) | `claude -p`, GitHub Actions or GitLab CI/CD ([Anthropic](https://code.claude.com/docs/en/overview)) |
| Models | Cursor's Composer and Grok models, plus Claude, GPT and Gemini ([Cursor](https://cursor.com/docs/models-and-pricing)) | Claude models only, through a Claude plan, an API key, or Amazon Bedrock, Google Cloud or Microsoft Foundry ([Anthropic](https://code.claude.com/docs/en/overview)) |

## Plans and prices

Both companies sell plans at $100 and $200 a month, which makes them look alike. The difference is what the usage covers. A Cursor plan includes a weekly pool for Cursor's own models. A Claude plan includes usage of Claude models, shared between Claude Code and Claude chat.

Cursor and Claude plans at each monthly price. Individual plans, from cursor.com/pricing and claude.com/pricing on November 9, 2026.

| Monthly price | Cursor | Claude Code |
| --- | --- | --- |
| $0 | Free: Limited usage | Free: Claude Code not included |
| $8 | Plus: Weekly pool for Composer and Grok; no Claude models | No plan at this price |
| $20 | No plan at this price | Pro: Five-hour and weekly limits; Claude models included; cloud agents |
| $30 | Premium: 4× the Plus pool; Claude models at API price; cloud agents | No plan at this price |
| $100 | Super: 15× the Plus pool; Claude models at API price; cloud agents | Max: 5× Pro usage; Claude models included; cloud agents |
| $200 | Ultra: 50× the Plus pool; Claude models at API price; cloud agents | Ultra: 10× Pro each week, no five-hour limit; Claude models included; cloud agents |

Neither company says how many tokens a plan includes. Cursor describes its plans as multiples of Plus, and Anthropic describes its plans as multiples of Pro, so the multiples can't be compared across the two. Team plans start at $30 a seat on Cursor and $25 a seat on Claude, or $20 billed annually ([Cursor](https://cursor.com/pricing), [Claude](https://claude.com/pricing)).

## What Claude models cost in Cursor

In Cursor, requests to Claude, GPT and Gemini are never part of the weekly pool. On Premium and above they bill at the model's API list price from prepaid credit, starting with the first request ([Cursor](https://cursor.com/docs/models-and-pricing)). The chart shows how much that changes the price of a request.

Price per million tokens in Cursor, from Cursor’s models and pricing docs on November 10, 2026.

| Model | Input | Output | Paid from |
| --- | --- | --- | --- |
| Composer 2.5 | $0.50 | $2.50 | Weekly pool |
| Grok 4.7 | $2.00 | $6.00 | Weekly pool |
| Composer 2.5 (Fast) | $3.00 | $15.00 | Weekly pool |
| Claude Sonnet 5.5 | $2.00 | $10.00 | Prepaid credit, on top of the plan |
| Gemini 3.1 Pro | $2.00 | $12.00 | Prepaid credit, on top of the plan |
| Claude Opus 5.5 | $4.00 | $20.00 | Prepaid credit, on top of the plan |
| Claude Fable 5.2 | $8.00 | $40.00 | Prepaid credit, on top of the plan |

Output tokens from Claude Opus 5.5 cost 8 times as much as Composer 2.5, and Claude Fable 5.2 costs 16 times as much. Teams and Enterprise also pay a Cursor Token Rate of $0.25 per million tokens on most third-party models.

In Claude Code, a Claude plan covers Claude model usage until you reach a five-hour or weekly limit. You then wait for the reset, upgrade, or turn on usage credits billed at API rates ([Anthropic](https://support.claude.com/en/articles/12429409-extra-usage-for-paid-claude-plans)). So a team that wants Claude models all day will usually pay less with a Claude plan, and a team that is happy with Composer or Grok gets more from a Cursor plan. [Cursor pricing](/resources/cursor-pricing) covers the Cursor side in more detail.

## Cloud agents

Both products can run an agent in a cloud VM that keeps working after you close your laptop and returns a pull request. They differ in who can use them and where you can start them.

| | Cursor Cloud Agents | Claude Code cloud sessions |
|---|---|---|
| Plans | Premium and above, Teams and Enterprise ([Cursor](https://cursor.com/docs/cloud-agent)) | Pro, Max, Ultra, Team, and Enterprise premium seats ([Anthropic](https://code.claude.com/docs/en/claude-code-on-the-web)) |
| Runs on | A VM managed by Cursor. Self-Hosted Machines can run the agent's tools on your own hardware | A VM managed by Anthropic, or your own self-hosted environment |
| Start from | The web, desktop app, iOS app, Slack, Linear, GitHub or Bitbucket comments, and the API | The web, Claude mobile app, desktop app, `claude --cloud` and scheduled tasks |
| Git hosts | GitHub and Bitbucket | GitHub. Other hosts can upload a local copy, but the session can't push back |
| Billing | Counts against the plan at the selected model's rate | Counts against the plan's usage limits, with no charge for the VM |
| Amazon Bedrock or Google Cloud | Not documented ([Cursor](https://cursor.com/docs/customizing/aws-bedrock)) | Not available, because cloud sessions need a claude.ai sign-in |

## Where Replicas fits

Replicas, which publishes this guide, runs Claude Code, Cursor, Codex, OpenCode and Pi in cloud workspaces. Each task gets its own Linux VM built from an environment your team defines once, and you can start tasks from the web, desktop or iOS app, Slack, Linear, GitHub or GitLab.

Claude Code signs in with your Claude subscription, and Cursor uses a Cursor API key that bills to your Cursor plan ([Replicas](https://docs.replicas.dev/features/coding-agents/cursor)). Replicas charges for workspace time and doesn't resell model usage. An admin can set a fallback chain, so a chat that reaches a Claude usage limit moves to another credential or another agent with the conversation carried over ([Replicas](https://docs.replicas.dev/admin/credentials#fallback-chains)).

Related comparisons: [Codex vs Claude Code](/resources/codex-vs-claude-code), [Replicas vs Cursor](/resources/replicas-vs-cursor) and [Replicas vs Claude Code](/resources/replicas-vs-claude-code).