QUICK ANSWER
What are the best vibe coding tools?
There is no single best vibe coding tool, because the nine here do different jobs. Cursor, Windsurf and Claude Code edit code on a developer machine. Bolt, Lovable, Base44, Replit and Playcode build and also run an app for you. v0 sits in between. Pick by what has to be running afterwards, then check the meter that bills you.
Most vibe coding lists rank tools that are not competing for the same job. An editor that edits files on your laptop and a browser product that builds and hosts an app for you are not alternatives to each other, and a ranking that mixes them tells you nothing useful about either.
So this guide sorts nine tools by what you are left holding when the session ends, then compares them on six things a buyer can actually check: who it is for, what it produces, whether data has somewhere to live, whether it publishes, whether you can take the project away, and what meters the bill. Every finding is dated first-party documentation, gathered on 16 August 2026. Nothing here was run as a pilot, and there is no overall winner at the end.
How this comparison was built
Five rules, applied before any brand was looked at. They exist to stop the two failure modes of a vendor-published comparison: a rigged rubric, and a feature table that reads like fact but was never checked.
Fix the definition first, so the list is not arbitrary
Vibe coding, in its original sense, means building software by describing it and accepting the code without reviewing it closely. That definition covers both the editors developers drive and the browser products a non-developer drives, which is why both are in this list. It also explains why they cannot be ranked against each other on one axis.
Sources: [karpathy-definition]
Sort by what is left running, not by feature count
The first split is whether the tool leaves you with edited files on your own machine or with a running application on a URL. That single question changes who can use the tool, what the tool costs to keep alive, and who fixes it at the weekend.
Sources: [claude-code-product], [bolt-home], [lovable-hosting]
Read the meter before the price
Every product in the hosted group and most of the editor group meter usage: credits, tokens, messages, or web requests. We recorded the entry price and the meter together, because the plan price alone does not predict a heavy month.
Sources: [cursor-pricing], [v0-pricing], [replit-pricing], [base44-pricing], [bolt-pricing]
Ask the exit question at the start
For each tool we looked for a documented way to get the project out in a runnable form. Where the retrieved pages did not answer it, the cell says the source did not establish it, which is a research gap on our side and not a claim about the product.
Sources: [playcode-cloud], [claude-pricing], [lovable-publish]
Record what could not be retrieved instead of filling it in
One vendor was unreachable to us on the day. Both of its documented URLs now resolve into Devin, which is itself the most useful thing a buyer can learn about it, so we published the redirect and the gap rather than a plausible-looking row.
Sources: [windsurf-pricing-redirect], [windsurf-docs-devin], [playcode-pricing]
What this guide covers, and what it does not
Read the exclusions before the table. Several of them are the reason a tool that looks wrong for you here may still be the right choice.
Included
- Nine tools that a person can use to build working software by describing it, spanning both developer editors and browser builders.
- Six buyer-checkable criteria, each answered from the vendor's own current documentation with the URL and the date attached.
- The entry price and the usage meter for each tool, stated together.
- An explicit gap wherever a retrieved page did not answer a criterion.
Not included
- Output quality. We did not build the same app nine times, so nothing here says whose generated code is better.
- Support responsiveness, uptime, and reliability. Asserting a rival's reliability without measuring it would be the same error as inventing a feature.
- Enterprise procurement detail: SSO, audit logs, compliance certifications, and contract terms.
- Native mobile output, except where a vendor's own documentation states its scope plainly.
- Any overall ranking or score. The rubric deliberately produces a shortlist per job, not a winner.
The six criteria, and why each one is on the list
Each criterion is something a buyer can verify on a vendor page in a few minutes, before spending anything. Feature counts, model names, and benchmark scores are deliberately absent: they change monthly and none of them predict whether the tool fits the job.
Who it is for, and where you work
Some of these are editors and terminal agents that assume you already have a codebase, a machine set up, and the ability to read what changed. Others are browser products that assume you have neither. Buying across that line is the single most common mistake.
What it actually produces
A pull request, a React component, a whole web app, or an app plus the machine it runs on are four different outputs. The question is not what the tool can generate but what you are left holding when it stops.
Backend, database, and stored data
The moment a project has logins, records, or uploads, it needs somewhere to keep them. Either the tool brings that, or you assemble and pay for it separately. This is where a weekend prototype stops being a weekend prototype.
Publishing, hosting, and custom domains
A working preview is not a live product. Check whether publishing is included, whether the live version updates automatically or only when you republish, and whether a custom domain needs a paid plan.
Code ownership and the exit path
Ask before you start, not after: can you get the project out, in a form somebody else can run? Editor-based tools work on files you already own. Browser builders vary, and the answer decides how expensive a change of mind is.
Entry price and what meters the usage
Almost none of these bill purely by seat. They bill by credits, tokens, messages, or requests, so the plan price is only half the number. The meter is what makes a busy month expensive.
The nine tools, in alphabetical order
Alphabetical order prevents an implied ranking. Playcode publishes this guide and appears in the same rubric as everything else, with the same source standard and its own limits stated in the row.
Base44
Best for: A non-developer who wants apps with logins and data, and prefers a credit allowance they can predict.
A browser AI app builder whose paid tiers are laid out as message-credit allowances, with backend, authentication, storage, and analytics documented as part of every plan.
| Criterion | Finding |
|---|---|
| Who it is for, and where you work | Documented: Positioned as a no-code AI platform for building apps, websites, and agents, with plan tiers from a free plan up to Elite. Nothing on the retrieved page assumes a local development setup. Sources: [base44-pricing] |
| What it actually produces | Documented: Apps, websites, and agents. The free tier is capped at up to 5 apps; the paid tiers are listed as an unlimited number of apps. Sources: [base44-pricing] |
| Backend, database, and stored data | Documented: Every plan is listed as including an integrated backend and database system, authentication and user management, and cloud storage. The free plan names authentication, database functionality, and analytics explicitly. Sources: [base44-pricing] |
| Publishing, hosting, and custom domains | Documented: Connecting a domain appears from the Starter tier upward, and the higher tiers add a free domain for one year. The free tier does not list domain connection. Sources: [base44-pricing] |
| Code ownership and the exit path | Not established by our sources: The pricing page lists in-app code edits from the Starter tier and a GitHub integration from the Builder tier, but does not describe a complete downloadable export. Check the current documentation before assuming either way. Sources: [base44-pricing] |
| Entry price and what meters the usage | Documented: Free at 25 message credits and 100 integration credits a month; Starter $16, Builder $40, Pro $80, and Elite $160 a month billed annually, carrying 100, 250, 500, and 1,200 monthly message credits respectively. Integration credits are metered separately. Sources: [base44-pricing] |
Tradeoffs
- Message credits are the unit of work, so a debugging-heavy week consumes the month faster than a calm one.
- A custom domain and in-app code edits sit above the free tier.
- The retrieved pricing page does not document a full project export, so the exit path needs checking before you commit.
Bolt
Best for: Someone who wants one browser product that generates the app and hosts it, and is comfortable with a token meter.
A StackBlitz browser builder that pairs generation with Bolt Cloud, described on its own site as hosting, databases, and integrations inside the same interface.
| Criterion | Finding |
|---|---|
| Who it is for, and where you work | Documented: Aimed at product builders working in the browser: "Create stunning apps & websites by chatting with AI." No local toolchain is implied anywhere on the retrieved pages. Sources: [bolt-home] |
| What it actually produces | Documented: Websites, apps, and prototypes, built by chatting, with the option to import an existing project. Sources: [bolt-home], [bolt-pricing] |
| Backend, database, and stored data | Documented: Bolt Cloud is described as backend infrastructure including hosting, databases, and integrations inside the same interface. Both the free and paid plans list unlimited databases; the paid plan adds expanded database capacity and a choice of database provider. Sources: [bolt-home], [bolt-pricing] |
| Publishing, hosting, and custom domains | Documented: Website hosting is on every plan, with up to 333,000 web requests on Free and up to 1 million on Pro. Bolt branding is removed and custom domain support appears on Pro. Sources: [bolt-pricing] |
| Code ownership and the exit path | Not established by our sources: Neither retrieved page documents a project export or a code download. The pricing page does mention private NPM registry support on the Teams plan, which implies real project files, but that is inference and not a documented export path. Sources: [bolt-pricing] |
| Entry price and what meters the usage | Documented: Free at 300,000 tokens a day and 1 million a month; Pro $25 a month starting at 10 million tokens a month with rollover of unused tokens; Teams $30 per member a month; Enterprise on quote. Yearly billing is advertised as saving up to 28 percent. Sources: [bolt-pricing] |
Tradeoffs
- Usage is metered in tokens with a daily cap on the free plan, which is a harder unit to estimate than messages.
- Free-plan sites carry Bolt branding, and custom domain support starts on the paid tier.
- The marketing claim of 98 percent fewer errors is the vendor's own and was not independently measured.
Claude Code
Best for: A developer or a technical team working inside an existing codebase from the terminal or an IDE.
Anthropic's agentic coding tool. It works on the repository you already have, runs commands, and opens pull requests. It does not host anything.
| Criterion | Finding |
|---|---|
| Who it is for, and where you work | Documented: Developers. The product page describes working directly in your codebase and shipping from the terminal, IDE, Slack, and web, available for macOS, Linux, and Windows. Sources: [claude-code-product] |
| What it actually produces | Documented: Changes to code you already have: multi-file edits, codebase explanations through agentic search, and complete issue-to-pull-request workflows integrated with GitHub and GitLab. Sources: [claude-code-product] |
| Backend, database, and stored data | Not established by our sources: No backend, database, or runtime is part of the product on the retrieved pages. It edits the code for whatever infrastructure you already run. Sources: [claude-code-product] |
| Publishing, hosting, and custom domains | Not established by our sources: No hosting, publishing, or domain handling is described. Deployment stays wherever your project already deploys. Sources: [claude-code-product] |
| Code ownership and the exit path | Documented: It edits files in your own repository and submits pull requests through your existing version control, so ownership and export are not separate questions. Sources: [claude-code-product] |
| Entry price and what meters the usage | Documented: Included in all paid Claude plans and sharing their usage limits: Pro at $17 a month with the annual discount ($200 billed up front) or $20 monthly, Max from $100 a month, Team at $20 per seat a month billed annually or $25 monthly. API access is priced per million tokens separately. Sources: [claude-pricing] |
Tradeoffs
- It assumes a codebase, a machine, and somebody who can judge a diff. It is the wrong purchase for a non-developer.
- Usage limits are shared across the whole Claude subscription, so heavy chat use and heavy coding use compete for the same allowance.
- Nothing about running the result is included; hosting, database, and domain remain your problem.
Cursor
Best for: A developer who wants an AI-first editor on their own machine, and a team that needs shared rules and admin controls.
The editor where the term vibe coding was born: Karpathy's original post named Cursor Composer as the tool he was using. It is a downloadable editor, not a hosting product.
| Criterion | Finding |
|---|---|
| Who it is for, and where you work | Documented: Individual developers and teams. The free tier is called Hobby and the pricing page leads with a download; Teams adds centralized billing, agentic code review, usage analytics, and SAML or OIDC single sign-on. Sources: [cursor-pricing] |
| What it actually produces | Documented: Code in the editor. Agent requests, Composer access, cloud agents, and MCP servers, skills, and hooks are listed as plan capabilities. Sources: [cursor-pricing] |
| Backend, database, and stored data | Not established by our sources: No backend or database is part of the product on the retrieved page. The editor writes code against whatever infrastructure you already have. Sources: [cursor-pricing] |
| Publishing, hosting, and custom domains | Not established by our sources: No hosting, publishing, or custom-domain handling appears on the pricing page. Deployment is out of scope. Sources: [cursor-pricing] |
| Code ownership and the exit path | Documented: It is a downloaded editor working on local files, so the project is already yours. Enterprise tiers add repository, model, and MCP access controls plus an AI code tracking API. Sources: [cursor-pricing] |
| Entry price and what meters the usage | Documented: Hobby is free with limited Agent requests and no credit card required. Individual is $20 a month with Pro, Pro+, and Ultra steps. Teams is $40 per user a month. Enterprise is custom with pooled usage. Bugbot is billed on usage. Sources: [cursor-pricing] |
Tradeoffs
- You still need a machine, a project, and the ability to read a diff.
- Agent request limits differ by tier, and some capabilities are billed on usage on top of the seat price.
- Nothing about running the finished app is included.
Lovable
Best for: A non-developer building a web app who wants publishing, access control, and a built-in backend without touching infrastructure.
A browser builder that hosts what it publishes, generates site metadata, and documents its scope clearly, including what it does not build.
| Criterion | Finding |
|---|---|
| Who it is for, and where you work | Documented: Teams and individuals building web apps in the browser. Roles matter: editors and above can publish by default, and Enterprise workspaces can restrict who may publish externally. Sources: [lovable-publish] |
| What it actually produces | Documented: Web apps that Lovable hosts. Its documentation is explicit about the boundary: "Lovable does not generate projects in React Native", and it recommends prototyping screens and flows in the browser then rebuilding them with React Native elsewhere. This is a scope statement, not a defect, and it is more than most vendors publish. Sources: [lovable-publish] |
| Backend, database, and stored data | Documented: A built-in backend, called Cloud, provides the database, authentication, storage, and functions behind the published pages, with a region setting controlling where backend data is stored. Usage-based run credits cover backend activity while the app runs. Sources: [lovable-hosting] |
| Publishing, hosting, and custom domains | Documented: Publishing to a lovable.app URL is free on all plans with HTTPS included and global delivery; custom domains require a paid plan and branded workspace URLs are Business and Enterprise. Only the current snapshot is live, so new changes need an explicit republish, and a project can be unpublished at any time. Sources: [lovable-hosting], [lovable-publish] |
| Code ownership and the exit path | Not established by our sources: The two retrieved pages address who can view a published site rather than how to take the project away. They state that publishing never exposes source code, chat history, or unpublished work, which is a privacy statement, not an export path. Check the current documentation. Sources: [lovable-publish], [lovable-hosting] |
| Entry price and what meters the usage | Not established by our sources: The plan ladder is Free, Pro, Business, and Enterprise, and usage is credit-metered, with metadata edits and publishing from chat both consuming credits. The prices themselves render client-side and did not appear in our fetch of the pricing page on 16 August 2026, so we quote no figure. Sources: [lovable-publish], [lovable-hosting] |
Tradeoffs
- Publishing is a snapshot: later changes are not live until you republish, which surprises people who expect a site to update itself.
- On Free and Pro, anyone with the link can visit a published app; restricting the audience requires Business or Enterprise.
- Its documentation states that it does not generate React Native projects, so a native mobile app means rebuilding outside Lovable.
Playcode
Best for: A non-developer who needs the finished thing to keep running: a real backend, a domain, and a way to undo a bad change.
The publisher of this guide. A browser AI builder with its own cloud runtime, where each project gets a machine with a database, HTTPS, background jobs, and whole-machine restore.
| Criterion | Finding |
|---|---|
| Who it is for, and where you work | Documented: Non-technical builders and small teams: business websites, MVPs, client portals, dashboards, internal tools, and small full-stack apps, with workspace billing, viewer, editor, and admin roles, and per-project guest sharing. Sources: [playcode-cloud], [playcode-pricing] |
| What it actually produces | Documented: Websites and full-stack web apps, plus the production machine they run on. Each project has a development machine for editing and a production machine that carries the domain, and publishing moves the exact commit between them. Sources: [playcode-cloud] |
| Backend, database, and stored data | Documented: A real database, durable file storage, background jobs, and HTTPS on a dedicated machine per project, with whole-machine restore that brings code, files, and database back together to a saved moment. Snapshots are taken automatically before risky changes, on request, and on demand. Sources: [playcode-cloud] |
| Publishing, hosting, and custom domains | Documented: Production hosting with a global CDN, automatic HTTPS, and custom domains with automatic SSL are listed on paid plans. Cloud bandwidth is 100 GB a month on Pro and 1 TB on Max. Sources: [playcode-pricing] |
| Code ownership and the exit path | Documented: Every project has its own git repository with full history that you can push and pull, and the whole project exports as a ZIP. Code export is listed as a paid-plan capability, so it is not available on the free tier. Sources: [playcode-cloud], [playcode-pricing] |
| Entry price and what meters the usage | Documented: Paid plans start at $21 a month with annual billing, or $25 monthly, and usage is metered in credits per message with credits rolling over: 100 a month on Pro and 400 to 10,000 on Max. Top-ups start at $25 and are valid for 12 months. The free tier is a trial with no cloud app and no custom domain. Sources: [playcode-pricing] |
Tradeoffs
- The free tier is a trial: full-stack cloud apps, custom domains, private projects, and code export all start on a paid plan.
- Cloud apps are capped by plan: 3 on Pro and 5 on Max, which is a real ceiling if you build many projects at once.
- Automatic scaling across machines and a two-way GitHub mirror are both stated as roadmap, not shipped.
Replit
Best for: Someone who wants a full development environment in the browser, with agents that can work in parallel.
A browser development platform with an Agent, a built-in database, publishing, and a credit allowance that scales steeply between tiers.
| Criterion | Finding |
|---|---|
| Who it is for, and where you work | Documented: A wide span, from a free Starter tier for exploring to a Pro tier described as being for commercial and professional builds, with collaborator limits of 5 on Core and 15 on Pro. Sources: [replit-pricing] |
| What it actually produces | Documented: Full-stack apps built with the Agent, plus slides, videos, and animations. Paid tiers allow multiple agents working in parallel: up to 2 on Core and up to 10 on Pro. Sources: [replit-pricing] |
| Backend, database, and stored data | Documented: A built-in database for full-stack apps is listed from the free Starter tier. Database rollbacks of up to 28 days appear on the Pro tier. Sources: [replit-pricing] |
| Publishing, hosting, and custom domains | Documented: Starter publishes up to 1 project and supports private or password-protected deployments. Core adds publishing in any region, unlimited workspaces, and removal of the "Made with Replit" badge. Sources: [replit-pricing] |
| Code ownership and the exit path | Not established by our sources: The retrieved pricing page does not describe export or download. It is a full development environment, so files exist, but no documented exit path appeared in our fetch. Sources: [replit-pricing] |
| Entry price and what meters the usage | Documented: Starter is free with daily Agent credits. Core is $25 a month, or $20 billed annually, and carries $25 of monthly credits. Pro is $100 a month, or $95 billed annually, with $100 of monthly credits. Enterprise is a sales conversation, and effort-based pay-as-you-go is offered on top. Sources: [replit-pricing] |
Tradeoffs
- The free Starter tier publishes at most one project, so a second live idea means paying.
- Credits are the unit and the jump from Core to Pro is four times the price, which suits professional use more than casual use.
- The retrieved pricing page does not document a project export path.
v0
Best for: A team already on Vercel that wants to generate interfaces and full-stack apps and deploy them into that stack.
Vercel's AI builder, sitting between the editor group and the hosted-builder group: it generates and deploys, and it exposes model pricing openly.
| Criterion | Finding |
|---|---|
| Who it is for, and where you work | Documented: Individuals and teams building for the web, with collaboration, shared chats, centralized billing on vercel.com, and enterprise options including SAML single sign-on and role-based access control. Sources: [v0-pricing] |
| What it actually produces | Documented: Full-stack web applications, described as designing, iterating, and scaling them, with a visual Design Mode for editing. Sources: [v0-pricing] |
| Backend, database, and stored data | Not established by our sources: The pricing page speaks of full-stack applications and deployment to Vercel but does not itemize a bundled database or authentication layer. Data services would come from the surrounding platform. Sources: [v0-pricing] |
| Publishing, hosting, and custom domains | Documented: Deploying apps to Vercel is listed from the free tier. Custom domain handling belongs to the Vercel platform rather than to v0 itself on this page. Sources: [v0-pricing] |
| Code ownership and the exit path | Documented: GitHub sync is listed from the free tier, which puts the generated project in a repository you control. Sources: [v0-pricing] |
| Entry price and what meters the usage | Documented: Free with $5 of monthly credits and a 7 message per day limit. Plus is $30 per user a month with $30 of credits, Business is $100 per user a month with the same $30 of credits plus training opt-out by default, and Enterprise is custom. Models are separately priced per million tokens, from $0.20 input on v0 Mini to $10 input on v0 Max Fast. Sources: [v0-pricing] |
Tradeoffs
- The free tier caps messages per day and includes only a small credit allowance.
- Its natural home is the Vercel platform, so it fits an existing stack more than a standalone one.
- Model choice changes the per-token price by up to fifty times, so the same work can cost very differently.
Windsurf
Best for: A developer evaluating the Cognition and Devin family, who should confirm the current product boundaries directly.
Now folded into Devin. On 16 August 2026 both Windsurf URLs we tried resolved into Devin properties, and the pricing page was not retrievable behind a bot checkpoint.
| Criterion | Finding |
|---|---|
| Who it is for, and where you work | Documented: Developers. The Windsurf getting-started documentation URL now serves "Welcome to Devin Desktop", described as a next-generation AI IDE with installers for macOS, Windows, and Linux that can import VS Code or Cursor settings. Sources: [windsurf-docs-devin] |
| What it actually produces | Documented: Code, in a desktop IDE on your own machine. As with the other editor-class tools, nothing about running the finished application is included. Sources: [windsurf-docs-devin] |
| Backend, database, and stored data | Not retrievable: Not answered by the pages we could reach on 16 August 2026. Sources: [windsurf-pricing-redirect], [windsurf-docs-devin] |
| Publishing, hosting, and custom domains | Not retrievable: Not answered by the pages we could reach on 16 August 2026. A desktop IDE would not normally include hosting, but we did not verify that. Sources: [windsurf-pricing-redirect], [windsurf-docs-devin] |
| Code ownership and the exit path | Documented: A desktop IDE operating on local files, with documented installers for three operating systems, so the project sits on your own machine. Sources: [windsurf-docs-devin] |
| Entry price and what meters the usage | Not retrievable: A request to windsurf.com/pricing on 16 August 2026 followed a redirect to devin.ai/pricing and returned HTTP 429 behind a bot checkpoint. No plan or meter detail was obtained, so none is quoted. Sources: [windsurf-pricing-redirect] |
Tradeoffs
- The product identity moved. Anything written about Windsurf as a standalone product needs re-checking against Devin's current documentation.
- We could not retrieve pricing, so no figure is quoted here at all.
- Everything below is thinner than the other eight rows, and that gap is the finding rather than a judgement.
Five rules that pick the shortlist for you
Apply them in order and the nine collapse to two or three. Every rule names what you give up, because a rule without a tradeoff is an advertisement.
You cannot read code, and the result has to be live for real customers.
Choose: Shortlist the hosted builders that document a backend and publishing in one product: Base44, Bolt, Lovable, Playcode, or Replit. Rule out the editor-class tools entirely.
Tradeoff: You accept the vendor's runtime and its usage meter, and the exit path becomes a question you must answer before you build, not after.
You already have a codebase and somebody who can review a diff.
Choose: Shortlist Claude Code, Cursor, or the Devin family that Windsurf now belongs to. They work on the repository you have, and the review step is the point.
Tradeoff: Nothing about running the software is included. Hosting, database, domain, and deployment stay exactly as much work as they were.
The app must store customer records, logins, or uploads from day one.
Choose: Restrict the shortlist to tools whose own documentation names a database: Base44, Bolt, Lovable, Playcode, and Replit all do. Confirm where the data is stored and what happens on restore.
Tradeoff: A bundled backend is faster to start and harder to move. Ask what a migration looks like before the data becomes valuable.
You are testing an idea this week and may throw it away on Friday.
Choose: Use a free tier and ignore the ceiling. Cursor Hobby, Bolt Free, v0 Free, Base44 Free, and Replit Starter all let you produce something without paying.
Tradeoff: Free tiers carry branding, publish limits, daily caps, or all three. They tell you whether the tool suits you, not whether it suits the finished product.
The project will be handed to a developer, an agency, or a buyer later.
Choose: Make the exit path a gate, not a footnote. Ask for a runnable export or a repository you control, and confirm it in a trial before the project has value.
Tradeoff: This narrows the field early and can eliminate a tool that would have been faster for the first two weeks.
Disclosure
Playcode publishes this guide, and is one of the nine rows
We are not neutral about our own product, so we held it to the same rule as everything else: only what our published pages state, with the limits named in the row. If the thing that has to survive is the running app, its database, and the ability to undo a change, that is the row worth reading twice.
See Playcode vibe codingNo credit card required to start.
Playcode plan limits and credit meterThe figures in the Playcode row, on the page they come from.
What this guide cannot tell you
Every one of these is a real limit on how far the table should carry a decision. Read them before you treat any cell as settled.
- Nothing here was piloted. We did not build the same application in nine tools, so the guide compares documentation, not results.
- Prices and usage meters in this category change often. Every figure is dated 16 August 2026 and should be re-checked on the day you buy.
- Windsurf could not be retrieved. Its pricing URL redirected to Devin and returned a rate limit, so its row is deliberately thin.
- Lovable prices are missing because that page renders client-side and our fetch returned no figures. The absence is ours, not theirs.
- "Not established by our sources" means we did not find it on the pages we retrieved. It is not evidence that a capability is missing.
- Playcode publishes this guide. Treat the Playcode row as a disclosed interested party and verify it the same way you would verify any vendor claim.
- Quality of generated code, support, and reliability are all excluded. They matter enormously and none of them can be established from a pricing page.
Source ledger
Fifteen first-party pages, each fetched on 16 August 2026. Every finding above cites at least one of these by ID, and nothing in the table comes from a source that is not listed here.
[karpathy-definition] Simon Willison:Not all AI-assisted programming is vibe coding (but vibe coding rocks)
Checked August 16, 2026. Supports: Quotes Andrej Karpathy's February 2025 post in full, including the reference to Cursor Composer, and defines vibe coding as building software with a model without reviewing the code.
[base44-pricing] Base44:Base44 Plans & Pricing
Checked August 16, 2026. Supports: Plan ladder and monthly message and integration credit allowances, app limits, domain connection by tier, GitHub integration by tier, and the list of what every plan includes.
[bolt-pricing] Bolt (StackBlitz):Plans & pricing
Checked August 16, 2026. Supports: Token allowances and daily caps, plan prices, hosting and web request limits, database capacity, branding, custom domain support, and yearly billing discount.
[bolt-home] Bolt (StackBlitz):Bolt AI builder: websites, apps and prototypes
Checked August 16, 2026. Supports: Positioning as a chat-driven browser builder, project import, and the description of Bolt Cloud as hosting, databases, and integrations in one interface.
[claude-code-product] Anthropic:Claude Code product page
Checked August 16, 2026. Supports: Terminal, IDE, Slack and web surfaces, macOS, Linux and Windows availability, codebase understanding, multi-file edits, and issue-to-pull-request workflows with GitHub and GitLab.
[claude-pricing] Anthropic:Plans and pricing
Checked August 16, 2026. Supports: Free, Pro, Max, Team and Enterprise plan prices, and the statement that Claude Code is included in all paid plans and shares the same usage limits.
[cursor-pricing] Cursor:Cursor pricing
Checked August 16, 2026. Supports: Hobby, Individual, Teams and Enterprise tiers with their prices, Agent request limits, Composer access, cloud agents, usage-based Bugbot billing, and team administration features.
[lovable-publish] Lovable:Publish your Lovable project
Checked August 16, 2026. Supports: Snapshot publishing and republishing, plan-dependent website access control, publish permissions, generated site metadata, unpublishing, and the statement that Lovable does not generate React Native projects.
[lovable-hosting] Lovable:How Lovable hosts your app
Checked August 16, 2026. Supports: Free lovable.app publishing with HTTPS and global delivery, paid-plan custom domains, run credits for backend activity, and the built-in Cloud backend covering database, authentication, storage and functions.
[playcode-pricing] Playcode:Pricing - Playcode AI and Cloud
Checked August 16, 2026. Supports: Free, Pro and Max tiers, monthly credit allowances and rollover, cloud app counts, bandwidth, custom domains, code export, and top-up validity.
[playcode-cloud] Playcode:Playcode Cloud: backend, database and hosting
Checked August 16, 2026. Supports: Dedicated machine per project, database and durable storage, HTTPS and custom domains, whole-machine restore of code, files and database, per-project git repository and ZIP export, and the roadmap status of cross-machine scaling.
[replit-pricing] Replit:Replit pricing
Checked August 16, 2026. Supports: Starter, Core, Pro and Enterprise tiers with monthly and annual prices, credit allowances, collaborator and parallel-agent limits, publish limits, database rollback window, and effort-based pay-as-you-go.
[v0-pricing] Vercel:v0 plans and pricing
Checked August 16, 2026. Supports: Free, Plus, Business and Enterprise tiers, included credits, daily message limit, GitHub sync, deployment to Vercel, Design Mode, and per-million-token model pricing.
[windsurf-pricing-redirect] Windsurf (Cognition):Windsurf pricing URL
Checked August 16, 2026. Supports: An HTTP request on 16 August 2026 followed a redirect to devin.ai/pricing and returned HTTP 429 behind a bot checkpoint. No plan detail was retrieved.
[windsurf-docs-devin] Windsurf (Cognition):Windsurf getting-started documentation URL
Checked August 16, 2026. Supports: The URL serves "Welcome to Devin Desktop", a downloadable AI IDE for macOS, Windows and Linux that can import VS Code or Cursor settings.
Questions buyers ask about vibe coding tools
What is the best vibe coding tool for a beginner?
A browser tool that also runs what it builds, so there is no local setup and no separate hosting decision. Base44, Bolt, Lovable, Playcode, and Replit all document that shape. Cursor, Claude Code, and the Devin family that Windsurf joined assume you already have a codebase and can read a diff, which makes them the wrong first purchase even though they are excellent tools.
Do I have to pay to try vibe coding?
No. On 16 August 2026, Cursor Hobby, Bolt Free, v0 Free, Base44 Free, and Replit Starter all documented a free tier, and Playcode offers a trial with credits included and no credit card required. Free tiers carry real limits: v0 caps messages at 7 a day, Bolt caps tokens daily and adds branding, and Replit Starter publishes at most one project.
Which vibe coding tools include a database?
Base44 lists an integrated backend and database on every plan. Bolt describes unlimited databases through Bolt Cloud. Lovable documents a built-in Cloud backend with database, authentication, storage, and functions. Replit lists a built-in database from its free tier. Playcode gives each project a machine with its own database. Cursor and Claude Code include none, by design.
What is the difference between a vibe coding tool and an AI code editor?
An AI code editor edits files on a machine you control and leaves running the software to you. A hosted vibe coding tool builds the app and keeps it running on a URL. The confusion is expensive in both directions: a non-developer buying an editor gets code with nowhere to live, and a developer buying a hosted builder pays for a runtime they already have.
Can these tools build a native mobile app?
Scope differs and is worth checking per tool. Lovable states it plainly in its own documentation: it does not generate React Native projects, and it suggests prototyping the screens in the browser and rebuilding them with React Native elsewhere. We did not verify native mobile output for the other eight on 16 August 2026, so treat that question as open and check the vendor documentation.
How much do vibe coding tools cost per month?
Entry paid plans clustered between roughly $16 and $30 a month on 16 August 2026: Base44 Starter $16, Cursor Individual $20, Playcode from $21 annually or $25 monthly, Bolt Pro $25, Replit Core $25, and v0 Plus $30 per user. Professional tiers rise steeply, to $100 a month at Replit Pro and v0 Business. Almost every plan also meters usage, so the plan price is a floor.
Why is Playcode in a list that Playcode publishes?
Because leaving it out would be its own kind of dishonesty. It is disclosed at the top of the table, ordered alphabetically like everything else, held to the same source standard, and its limits are stated in its own row: no cloud app or export on the free tier, a cap of 3 to 5 cloud apps by plan, and cross-machine scaling still on the roadmap.
If the running part is the hard part
Describe the app. Playcode builds it and runs it.
A real backend and database, hosting with HTTPS and a custom domain, and a restore that brings code, files, and the database back to a saved moment together. Everything in the Playcode row above, on the page that documents it.
Start BuildingNo credit card required. AI credits included to start.