Claude in VSCode Pricing | Comparing Pro/Max/API Plans and Monthly Costs
For developers who have paused their subscriptions because they aren't sure how much it costs to use Claude Code from VSCode each month, this article organizes Anthropic's official pricing across 6 tiers — Pro, Max 5x, Max 20x, Team, Enterprise, and API pay-as-you-go — covering annual discount rates, why there's no free option, and the minimum plan you should start with, all based on official documentation.
The VSCode Claude Code extension itself is completely free and can be installed in one click from the Marketplace at vscode:extension/anthropic.claude-code. However, as the official setup documentation clearly states, "Claude Code requires a Pro, Max, Team, Enterprise, or Console account," meaning Claude Code will not launch with a free Claude.ai plan. At minimum, a Pro plan is required.
The most affordable option for individual developers is Pro at $17/month with annual billing ($20/month with monthly billing), which includes Claude Code. If you hit usage limits with heavy use, the next step up is Max 5x at $100/month or Max 20x at $200/month (both monthly), and if that's still not enough, switching to API pay-as-you-go (Console account) is the officially recommended progression.
For team adoption, Team Standard at $20/seat/month (annual) or Premium at $100/seat/month are available, as well as Enterprise ($20/seat + API usage fees) for larger organizations. If your organization is already on AWS, GCP, or Azure, you can also run the same extension via Bedrock, Vertex AI, or Microsoft Foundry to consolidate billing.
目次 (8)
- Conclusion: The VSCode Extension Is Free — Only the Claude Subscription Costs Money
- Pro Plan: The Minimum for Individual Developers (Annual: $17/month, Monthly: $20/month)
- Max 5x / Max 20x: Raised Limits for Heavy Users ($100–$200/month)
- Team Plan: The Standard for Team Adoption (Standard $20/seat and up, Premium $100/seat and up)
- API Pay-as-You-Go (Console Account): Using the VSCode Extension with an API Key
- Bedrock / Vertex AI / Microsoft Foundry: Consolidating Billing Through Your Cloud Provider
- Choosing the Right Plan: 3 Decision Points
- Sources
Conclusion: The VSCode Extension Is Free — Only the Claude Subscription Costs Money
The first thing to understand is that the Claude Code extension for VSCode is distributed for free on the official Anthropic Marketplace. You can search for "Claude Code" in the Extensions panel and install it with one click, and the same extension works in VSCode forks like Cursor, Windsurf, and Kiro (VS Code official documentation).
The cost comes not from the extension itself, but from your Claude subscription or API usage. When you first launch the extension after installation, it opens a browser sign-in, and the account types accepted here are officially limited. Quoting the relevant section of the setup documentation:
Claude Code requires a Pro, Max, Team, Enterprise, or Console account. The free Claude.ai plan does not include Claude Code access.
In other words, even if you install the VSCode extension with a free Claude.ai account, Claude Code will not start (Setup documentation). Most cases of "I installed the extension but it won't work" are caused by getting stuck at this authentication step. The discussion of pricing starts from the premise that "at minimum, a Pro or higher subscription is required."
Note that if you configure the setup to call the Anthropic API via Bedrock, Vertex AI, or Microsoft Foundry, you can skip a Claude subscription entirely and consolidate billing through your cloud provider (covered below).
Pro Plan: The Minimum for Individual Developers (Annual: $17/month, Monthly: $20/month)
The Pro plan is the most affordable way for individual developers to use Claude Code from VSCode. According to the official pricing page, the cost is $17/month with annual billing or $20/month with monthly billing (as of May 2026), and all paid plans at Pro level or above include the "Includes Claude Code" label. For a deeper dive into pricing across all plans, see the Complete Guide to Claude Pricing.
The Pro plan gives you access to Claude.ai on Web, iOS, and Android, plus desktop Claude Code (CLI + VSCode extension), the JetBrains plugin, and Claude Code on the web. Within VSCode, you can use the chat panel in the right pane, launch from the Spark icon, specify line ranges with @file.ts#5-10, use Plan mode, preview diffs before applying, and roll back with checkpoints — all available on Pro.
One important caveat: the official pricing page explicitly states "Usage limits apply," and Pro includes usage caps per 5-hour window. Pro is sufficient for a few hours of code generation per day with Sonnet-class models, but frequent use in long pair-programming sessions or parallel sessions may result in hitting limits often. From there, upgrading to the Max plan or using API pay-as-you-go alongside is the officially recommended path (for details on how limits work and how to handle errors, see Claude Usage Limits and Error Handling).
Max 5x / Max 20x: Raised Limits for Heavy Users ($100–$200/month)
The next tier on the official pricing page is the Max plan, described as "$100/month and up" in a tiered structure. Specifically, Max 5x is $100/month and Max 20x is $200/month (both monthly rates), where the number represents the multiplier on usage limits compared to Pro. While Pro had two billing options (annual at $17/month, monthly at $20/month), Max is listed on the official pricing page as "$100/month and up" with only monthly pricing shown — there is no two-tier "annual/monthly" display or explicit annual discount like Pro. Therefore, the $100 and $200 figures for Max 5x and Max 20x should be understood as monthly billing amounts (the same even with annual billing), not annual effective rates.
The Max plan targets use cases like "running Claude Code for several hours every day," "running parallel sessions across multiple VSCode windows," or "using the Opus model for deep reasoning." From within the VSCode extension, you can check plan usage with the /usage command, and official features like switching down to Sonnet before hitting limits, or compressing context with /compact, are available as cost-saving measures.
Note that even Max 20x does not mean unlimited usage. Anthropic's official FAQ clearly states a "fair use" policy, and extreme parallel execution may be subject to restrictions. If Max 20x still isn't enough, or if parallelism is business-critical, API pay-as-you-go (described below) is the workaround.
Max 5x and Max 20x support mid-month upgrades and downgrades, with prorated billing applied when switching during the month. Flexible usage like "Max 20x for just the two weeks before next month's release" is a practical option.
Team Plan: The Standard for Team Adoption (Standard $20/seat and up, Premium $100/seat and up)
When sharing Claude Code in VSCode across multiple people, the Team plan is the standard choice. According to the official pricing, it comes in two tiers: Team Standard at $20/seat/month (annual) / $25/seat/month (monthly), and Team Premium at $100/seat/month (annual) / $125/seat/month (monthly).
The main differences between Standard and Premium are usage limits and the depth of management features. Premium offers higher usage limits comparable to Max, plus team management features like SSO (Single Sign-On, which lets employees log into services with their company identity), SCIM (a protocol for automatically syncing account information and provisioning with your identity platform), audit logs, and usage analytics. From the VSCode extension, signing in with a Team account brings up the same UI as individual Pro, and sharing .mcp.json via a Git repository allows team-wide distribution of MCP (Model Context Protocol — the standard for connecting external tools and data to Claude; see the What is MCP explainer) server settings.
Since this is per-seat billing, calculate based on the number of developers who actually use Claude Code daily. Teams that have "contracted 20 seats but only 5 people actually use it" may save money by switching to 5 Premium seats. Billing supports both credit cards and annual lump-sum transfers.
API Pay-as-You-Go (Console Account): Using the VSCode Extension with an API Key
There is also a path to running the VSCode extension with token-based pay-as-you-go rather than a subscription. Create an account in the Anthropic Console, then pass ANTHROPIC_API_KEY via an environment variable or through ~/.claude/settings.json, and the VSCode extension starts in API key authentication mode.
The benefits are three: "pay only for what you use," "no concept of parallel limits (though per-model RPM/TPM limits exist — RPM = requests per minute, TPM = tokens per minute)," and "billing can be consolidated under an organizational ID by department." The downside is unpredictability — running a long Opus refactoring session can easily cost several dollars in a single session. The standard approach to cost control is to set a hard cap with the Console's Spend Limits feature.
As a rough cost reference: with Sonnet 4-class models, expect "a few dozen to a few hundred yen per session"; with Opus, "a few hundred to a few thousand yen per session" (varies significantly based on prompt length and diff volume). Developers who prefer predictable monthly budgets should choose Pro/Max, while those with highly variable monthly workloads may find API pay-as-you-go safer for actual-cost billing.
Note that if you hold both a Pro/Max subscription and an API key, you can manually switch to API mode the moment you hit the subscription's rate limit.
Bedrock / Vertex AI / Microsoft Foundry: Consolidating Billing Through Your Cloud Provider
For organizations already committed to AWS, GCP, or Azure that don't want to add separate Anthropic contracts, cloud-provider access is an option. By enabling disableLoginPrompt in the VSCode extension and adding provider settings to ~/.claude/settings.json, the same extension switches to a mode where it calls Claude via Amazon Bedrock, Google Vertex AI, or Microsoft Foundry (VS Code official documentation).
Pricing follows each cloud provider's "Claude model pay-as-you-go" table, which is roughly on par with Anthropic's direct API rates. The key differentiators are unified billing and streamlined contracts and compliance — benefits like AWS Enterprise Discount Program pricing, Google Cloud committed use discounts, and Microsoft Azure EA pricing can all be applied to Claude as well.
There are two caveats. First, some subscription features of the VSCode extension (such as Remote session restoration via Claude.ai sign-in) may be unavailable. Second, cloud provider availability of model regions and SKUs may lag slightly behind direct Anthropic access — for example, the latest Opus may be available on direct contracts but not yet on cloud versions for a period. A simple rule of thumb: prioritize Anthropic direct for speed, cloud-provider access for governance.
Choosing the Right Plan: 3 Decision Points
To wrap up, here are 3 key points for choosing a pricing plan when using Claude Code in VSCode. The first is "individual or team?" — if it's just you, the choice is among Pro, Max, and API; for a team, the choice is among Team, Enterprise, and cloud-provider access.
The second is "can you predict your monthly usage hours?" If you can, the fixed monthly cost of Pro or Max is easier to budget; if not, API pay-as-you-go with actual-cost billing is safer. The third is "what cloud contracts does your organization already have?" If there are large existing contracts with AWS, GCP, or Azure, cloud-provider access should be your first consideration.
In terms of concrete pricing, the most common path for individual developers is: start with Pro annual billing at $17/month, then upgrade to Max 5x if you hit usage limits. For team adoption, starting small with Team Standard at $20/seat/month and switching to Premium or Enterprise as confidentiality requirements or team size grows is the realistic progression. Since the VSCode extension itself is free and subscription plans can be switched monthly, the fastest way to validate your needs is to start with Pro, install it, and see how quickly you hit the usage limits.