Vibe Coding Tools

What vibe coding really costs: subscriptions, tokens and hosting (2026)

A line-by-line look at what building with AI costs in November 2026: agent subscriptions, usage-based model pricing, app builder credits, hosting and databases, with worked monthly budgets for three typical setups.

Floot ·

Key Takeaways

  • There are three cost buckets, not one. What you pay to build (agent or builder plans), what model usage costs beyond the plan, and what the live app costs to run (hosting, database, email, domains).
  • Subscriptions are cheapest for steady daily use. Claude Pro and ChatGPT Plus are $20 a month and include Claude Code and Codex respectively, within usage limits that reset on a schedule.
  • Usage-based pricing is where surprises happen. API tokens, credits and editor plans that bill some models at API rates can all run past what you expected on a heavy week.
  • Running costs start small and grow with users. A do-it-yourself stack of Supabase Pro and Vercel Pro is $45 a month at list prices before usage. All-in-one platforms bundle some hosting into the plan.
  • Set limits before you need them. Spend caps, usage alerts and a monthly review of every bill prevent most cost shocks.

Quick-Start Prioritization Framework

SetupBuildingRunning the appExample monthly total (list prices)
Hobby: agent subscription + free tiers$20$0 (free tiers, non-commercial)$20
Founder, wired yourself$20–$100$45 (Supabase Pro + Vercel Pro)$65–$145 plus usage
Founder, all-in-one builder$20–$100 builder planOften partly included$20–$100 plus usage
Heavy daily agent user$100–$200Varies$100–$200+ plus hosting

The headline price of an AI coding tool is rarely what you end up paying. A $20 subscription runs out mid-week and nudges you to a $100 tier. An editor’s cheap plan covers its own models but bills others at API rates. Then the app needs a database and hosting, and those have their own plans and usage meters. This guide breaks vibe coding costs into the three buckets that actually show up on statements (building, model usage and running the app), lists current list prices from each vendor’s own pricing page as of 11 November 2026, and works through three realistic monthly budgets. Prices in this market change often, so treat every figure as a snapshot and check the linked pages before you commit.

The three buckets

Every vibe coding bill falls into one of three buckets. Keeping them separate makes it obvious where to cut.

Building

The subscription for the tool you prompt: a coding agent plan (Claude, ChatGPT, Cursor, GitHub Copilot) or an app builder plan (Replit, Lovable, Bolt, v0 and others). This is mostly fixed per month.

Model usage beyond the plan

Most plans include a quota. Past it, you either wait for the reset, upgrade, or pay per use: API tokens, extra credits, or overage billing. This bucket is variable and is where most surprises come from.

Running the app

Hosting, the database, file storage, email sending, a domain, and any paid APIs the app itself calls (for example an AI feature inside your app). This bucket grows with users, not with how much you build.

Coding agent subscriptions

Prices below are monthly list prices from each vendor’s pricing page on 11 November 2026.

ProductEntry paid planHigher tiersNotes from the pricing page
Claude (includes Claude Code)Pro $20/mo ($17/mo billed annually)Max $100/mo (5x Pro usage); Ultra $200/moFree plan doesn’t include Claude Code
ChatGPT (includes Codex)Go $8/mo; Plus $20/moPro from $100/mo ($100, $200 or $500)Codex is included from Free upward; Pro “currently has no five-hour limit”
CursorPlus $8/mo (Grok models only)Premium $30, Super $100, Ultra $200Third-party models billed at API list price by usage on paid plans; allowances refresh weekly
GitHub CopilotPro $10/moPro+ $39, Max $100Plans include monthly AI credits; 1 credit = $0.01

How usage limits work

Agent subscriptions don’t sell a fixed number of prompts. Claude’s Max page describes usage as multiples of Pro “every 5 hours and every week.” OpenAI’s Codex pricing page gives estimated local message ranges per five-hour window for Plus and notes these “are not fixed message limits.” In practice, a long agent session with large files uses far more of the allowance than quick questions.

Which tier is right

If you build a few evenings a week, a $20 plan is usually enough. If you hit the limit most days and lose time waiting for resets, the $100 tier is the common next step. The $200 tiers are for people running agents for much of the working day.

Paying by the token

You can also use agents with an API key and pay per token. Claude Code accepts an Anthropic Console account, and OpenAI’s docs say Codex can run with an API key billed at standard API rates.

Current API prices

Anthropic’s pricing page lists, per million tokens: Claude Sonnet 5.5 at $2 input and $10 output; Claude Opus 5.5 at $4 input and $20 output; Claude Haiku 5.5 at $0.10 input and $0.50 output for prompts up to 100,000 tokens. Cache hits on Sonnet 5.5 cost $0.10 per million, which matters because agents resend the same context many times.

A worked example

Our own illustrative arithmetic, not a measured figure: suppose a session sends 3 million uncached input tokens and produces 300,000 output tokens on Sonnet 5.5. That’s 3 × $2 + 0.3 × $10 = $9. If 80% of the input were cache hits instead, input would cost 0.6 × $2 + 2.4 × $0.10 = $1.44, for about $4.44 in total (ignoring the smaller premium charged for writing the cache). Do that twenty days a month and the difference between cached and uncached work is roughly $90 versus $180.

When API billing makes sense

For automation (an agent in CI reviewing every pull request) or very irregular use. For daily interactive building, subscriptions are usually cheaper and, more importantly, capped.

App builder plans

All-in-one builders bundle the agent, and often hosting and a database, into credit-based plans.

BuilderEntry paid planWhat the credits cover
ReplitCore $20/mo ($18 billed annually)Includes $20 toward more powerful models; Pro is $100/mo with $100 toward the most powerful models and $100 in credits
LovablePro from $25/mo for 100 creditsCredit tiers up to 10,000; Lovable Cloud hosting draws on credits after a monthly grant
BoltPro $25/moStarts at 10M tokens a month, with hosting and databases
v0Plus $30/user/mo$30 of included monthly credits per user plus $2 free daily credits

The thing to watch

Credits measure AI work, and complex changes cost more than simple ones. A month of debugging a stubborn feature can use several times the credits of a month building straightforward pages. Check whether unused credits roll over: Bolt’s pricing page says Pro tokens carry forward, for example, while its Free tokens don’t.

Running the app: hosting and database

This bucket is separate from building, and it’s the one that keeps billing after you stop prompting.

Do-it-yourself stack

Supabase’s Free plan includes two active projects and a 500 MB database, but free projects pause after a week of inactivity. Pro starts at $25 a month with 8 GB of disk per project and daily backups kept for 7 days. Vercel’s Hobby plan is free but “for personal, non-commercial use”; Pro is $20 a month with $20 of included usage credit. For a commercial app that’s $45 a month before usage, as our own sum.

Bundled hosting

Builders and platforms often include some hosting in the plan and meter the rest. Read what’s included in money or requests, and what happens when you exceed it: an overage charge, a pause, or a hard stop.

The smaller line items

  • A domain: typically an annual fee at a registrar, sometimes included with a plan.
  • Transactional email: free tiers exist; volume costs extra.
  • AI features inside your app: if your app calls a model for users, you pay those tokens on every use.
  • Error tracking and uptime monitoring: free tiers cover most small apps.
  • Payment processing: a per-transaction fee charged by your payment provider.

Three worked monthly budgets

These use list prices from the pages cited and our own arithmetic. They exclude taxes and usage overages.

The weekend builder

Claude Pro or ChatGPT Plus at $20, building a personal tool on free hosting tiers. Total: $20 a month. Fine as long as the app isn’t commercial and you accept that free projects may pause when idle.

The founder with a first paying customer, wired yourself

Claude Pro $20, Supabase Pro $25, Vercel Pro $20: $65 a month. If the agent limit becomes a bottleneck, moving to Claude Max takes it to $145. Add a domain and email.

The heavy agent user

Someone running agents most of the day on a $200 tier, plus $45 of hosting and database: about $245 a month before usage. At this level, the hosting is the small part.

How to keep costs under control

A few habits prevent nearly all bill shocks.

  • Set spend caps and usage alerts wherever a vendor offers them, especially on API keys and usage-billed models.
  • Keep sessions focused. Starting fresh for each task reduces how much context is resent with every request.
  • Pick a cheaper model for simple work. Many tools let you switch models; quick edits rarely need the largest one.
  • Don’t pay for idle projects. Delete or downgrade apps nobody uses.
  • Review every bill monthly and write down what each one is for.
  • Read free-tier terms before going commercial; some exclude it.

Common mistakes

The cost errors that come up most often.

  • Comparing only the AI tool’s price and forgetting hosting, database and domain.
  • Leaving an API key with no spend limit in an automation that loops.
  • Choosing a cheap editor plan without noticing which models it bills at API rates.
  • Upgrading to a $200 tier before checking whether waiting for a reset would have been fine.
  • Adding an AI feature to your app without estimating per-user token cost.
  • Running a commercial app on a free tier whose terms say non-commercial.

If you’d rather have one bill

If you’d rather not juggle separate database and hosting bills, Floot is free to start, and its Pro plan is $25 a month with 1,000 building actions a day and $10 of hosting included each month; publishing is free on every plan and hosting beyond the allowance is metered by usage.

Frequently Asked Questions

How much does vibe coding cost per month?

From $20 a month for an agent subscription on free hosting, to roughly $65–$145 for a small commercial app on a do-it-yourself stack, to $200 or more for heavy daily agent use, at November 2026 list prices.

Is Claude Code free?

No. Claude’s pricing page lists Claude Code on Pro ($20 a month) and higher plans, not on Free. You can also use it with an Anthropic Console account and pay per token.

Is Codex included in ChatGPT Plus?

Yes. OpenAI’s Codex pricing page says Codex is included in Free, Go, Plus, Pro, Business, Edu and Enterprise plans, with usage limits that vary by plan.

Subscription or API key: which is cheaper?

For daily interactive building, subscriptions are usually cheaper and capped. API keys suit automation and irregular use, but set a spend limit.

What does hosting a vibe-coded app cost?

Free for non-commercial hobby use on many hosts. For a commercial app on Supabase Pro plus Vercel Pro, $45 a month at list prices before usage; bundled platforms vary.

Why did my AI coding bill spike?

Usually long sessions resending large context, an expensive model used for simple tasks, or usage billed at API rates beyond a plan’s allowance. Check the usage page of each tool.

Sources