If you've been trying to figure out which Claude Code plan actually makes sense for your business, you're not alone. The Anthropic pricing page gives you the numbers, but it doesn't tell you much about what those numbers mean in practice. When you'll hit limits, when the API route makes more sense, or whether the jump to Max is worth it.

We've navigated these tiers across real client builds at AMPL, from occasional business automation to heavy daily AI pipelines. This is what we've learned.

Claude Code pricing at a glance — the three tiers

There are essentially three ways to pay for Claude Code, and they work differently enough that picking the wrong one can either waste money or leave you hitting rate limits at the worst possible time.

Each tier suits a different kind of user. The right choice depends on how often you're using Claude Code, what you're building, and whether you're a single developer or a team running it across multiple projects.

Claude Code Pro ($20/month) — what you get and who it's for

Pro is the entry point. At $20 a month, you get access to Claude Code through the Claude.ai interface with usage limits that Anthropic describes as being on a rotating window, typically a five-hour reset cycle.

In practice, Pro is fine if you're using Claude Code to explore what it can do, run occasional automation tasks, or work on a single project a few times a week. If you sit down for an hour of focused work every few days, Pro handles it without complaint.

Where it starts to feel tight is when you're trying to run longer agentic tasks. Multi-step builds where Claude Code is working through a codebase, writing tests, fixing errors, and iterating. Those sessions use a lot of context, and they eat through your allowance faster than you'd expect.

Usage limits and what happens when you hit them

When you hit the Pro usage limit, Claude Code doesn't stop working entirely. You get a message telling you when your limit resets, usually within a few hours. It's not a hard block, but it does interrupt your flow if you're mid-build.

For business use, this is the main friction point with Pro. If you're using Claude Code as part of a daily workflow, running automation scripts, building integrations, iterating on live systems, hitting a limit at 2pm on a Tuesday is genuinely disruptive. That's usually the point where upgrading to Max starts making financial sense.

Claude Code Max ($100/$200/month) — when it's worth it

Max comes in two flavours: a $100/month plan with 5x the usage limits of Pro, and a $200/month plan with 20x. The step up from Pro to the lower Max tier is the most significant change in the lineup. You go from occasionally hitting walls to being able to work through a full day without thinking about limits.

For most developers and small teams running Claude Code seriously, the $100/month Max plan is the sweet spot. You're getting meaningful headroom without paying for capacity you won't use. The 20x tier at $200 is genuinely for heavy users. Think running multiple concurrent agentic sessions, working across large codebases, or being the person on a team who's running builds all day.

The difference between the two Max tiers

To be honest, most people won't need the $200 tier. If you're a solo developer or running Claude Code as part of a small team's automation work, the $100 plan gives you enough room to work freely throughout the day.

The $200 tier starts making sense when you're doing what I'd call infrastructure-scale automation. Building multiple systems simultaneously, running Claude Code agents that work unsupervised on long tasks, or when you're the technical lead on a team and Claude Code is open in your IDE for eight-plus hours a day.

At AMPL, we've found the $100 Max plan covers most of our project build work comfortably. The $200 tier comes into play during intensive phases, usually when we're deep in a complex integration build and context windows are running large.

Claude Code via API — pay-per-token explained

The API route works differently. Instead of a flat monthly fee, you're paying for what you actually use, measured in tokens (roughly 750 words per 1,000 tokens for context).

Current API pricing for Claude Sonnet (the model Claude Code defaults to for most tasks) sits at $3 per million input tokens and $15 per million output tokens. Claude Opus, the more capable model, is $15 input and $75 output per million tokens.

For production systems, where Claude Code is running as part of an automated pipeline rather than a live coding session, API pricing is usually the right model. You're not paying for idle time, and you can control costs precisely by managing what you send to the model.

When the API route saves you money (and when it doesn't)

The API makes economic sense in a few specific situations. If you're running automation at scale, say processing hundreds of documents a day or running Claude as part of a backend system, flat subscription pricing often works out more expensive than paying per token at that volume. You're only paying for what you use.

It also makes sense for teams. Instead of every developer buying their own Pro or Max subscription, you can route usage through a single API key, track it properly, and allocate costs to projects.

Where the API can catch you out is in agentic workflows with long contexts. When Claude Code is working through a large codebase, it's reading a lot of tokens on each step, and those input costs add up faster than people expect. I've seen projects where an agentic session that felt routine in the UI would have cost $15-20 via API because of the context being passed back and forth.

The practical rule: if you're doing interactive development work, subscription tiers are more predictable. If you're running production automation pipelines, API pricing is usually better. But set up token usage monitoring before you go live. Anthropic's dashboard shows usage breakdowns, and it's worth checking it regularly until you have a feel for your consumption patterns.

Claude Code for teams — what the business plan adds

Beyond the individual tiers, Anthropic offers a Teams plan at $30 per user per month with a minimum of five users. This sits on top of Claude's business features rather than being a direct comparison to the Pro/Max tiers.

The business plan adds things that matter if you're deploying Claude Code across a team: centralised billing, admin controls, SSO, and higher context windows. You also get priority access during high-demand periods, which matters more than it sounds if you're running time-sensitive automation builds.

For businesses running Claude Code as a core tool across multiple people, the Teams plan is worth considering. The admin visibility and centralised cost management alone justify the small premium over individual Pro subscriptions. It also gives you a single billing relationship with Anthropic rather than managing multiple personal accounts.

Which plan should you actually choose?

Here's how I'd think about it, based on what we've seen across different types of use.

Start with Pro if you're still working out whether Claude Code fits into your workflow. The $20 entry point is low enough that it's not a meaningful risk, and most people know within a couple of weeks whether they're hitting limits or not.

Move to Max ($100) when you're using Claude Code regularly. Meaning most working days, running sessions that last more than an hour, or when hitting a limit has actually disrupted your work once or twice. The jump from $20 to $100 feels significant, but if Claude Code is genuinely part of your workflow, it's a small cost relative to the time it saves.

Consider Max ($200) only if you're genuinely working at high volume all day. Most businesses don't need it.

Use the API if you're building production systems, need team-level cost visibility, or are running automated pipelines that don't require interactive sessions.

Decision matrix: use case vs plan

FAQ: Claude Code pricing questions

How much does Claude Code cost?

Claude Code starts at $20/month on the Pro plan, which suits occasional use. For regular daily use, Max plans cost $100 or $200/month depending on usage volume. API pricing is consumption-based, starting at $3 per million input tokens for Claude Sonnet, and is better suited to production pipelines than interactive development work.

What's the difference between Claude Code Pro and Max?

The main difference is usage limits. Pro gives you a rate-limited allowance that resets on a rolling cycle, fine for occasional work but disruptive if you're building daily. Max removes those interruptions with 5x or 20x the limits depending on which tier you choose. If Claude Code is genuinely part of your daily workflow, the jump to Max usually pays for itself in uninterrupted build time.

Can I use Claude Code on the API instead of a subscription?

Yes, and for production systems it's often the better choice. You pay per token rather than a flat fee, which works well for automated pipelines where you can predict and control usage. The risk is that agentic tasks with large contexts can run up costs faster than expected. Set up usage monitoring before you go live and check it regularly until you understand your consumption patterns.

Is the Claude Code Teams plan worth it for small businesses?

If you have five or more people using Claude Code regularly, yes. The centralised billing and admin controls are genuinely useful. You can see what's being used across the team, manage access properly, and avoid the mess of multiple individual subscriptions on different billing cycles. Below five people, individual Max plans usually make more sense.

What happens when you hit Claude Code usage limits?

On the Pro plan, you get a notification telling you when your limit resets, usually within a few hours. It's not a hard cutoff, but it does pause your work. On Max plans, you're less likely to hit limits during normal use. On the API, there are no usage limits as such, but you can set spend caps to prevent unexpected charges.

Does the plan affect which Claude model you can use?

Yes, to some extent. All plans give access to Claude Sonnet by default, which handles most coding and automation tasks well. Opus access depends on the plan tier, and via API you can choose the model explicitly and pay the corresponding per-token rate. For most business automation work, Sonnet is more than sufficient. The jump to Opus is usually only worth it for tasks requiring complex reasoning across very long contexts.

The bottom line

Claude Code pricing is straightforward once you understand what you're actually buying. Pro is a reasonable starting point. Max is for serious daily use. The API is for production systems where you need control over costs and scale.

The mistake most people make is staying on Pro too long because the monthly number feels lower, then spending time working around limits instead of working. If Claude Code is saving you meaningful hours, the $80 difference between Pro and Max pays for itself quickly.

If you're trying to work out whether Claude Code is the right tool for a specific business automation problem, we can help with that. Book a free audit at amplconsulting.ai and we'll map out exactly what makes sense for your situation.