Bolt.new review

Bolt.new is a hosted, browser-based AI builder for apps, websites, slides and prototypes. You describe what you want in a chat interface; Bolt routes tasks to different models and generates or changes the project. It is a SaaS workflow, rather than a library, CLI or self-hosted server you install in an existing codebase.

Bolt Cloud adds infrastructure around the generated project: hosting, databases, integrations, user management and authentication. The vendor also lists hosting analytics, custom domains and SEO optimization. That puts Bolt across several layers of a small web-app stack: code generation, backend services and deployment. You can begin from a prompt or start from Figma, GitHub or a team template. The appeal is fewer separate tools and accounts; the trade-off is that your build workflow and backend depend on Bolt's hosted environment. The homepage claims automatic testing and refactoring reduce errors by 98%, but gives no measurement method alongside that claim.

The main integration surface presented on the homepage is Bolt's chat-based builder, with Figma and GitHub listed as starting points, plus team templates. The site describes automatic model routing and Bolt Cloud services, but does not present a public API, SDK, programming-language list or CLI. Confirm SDK support and any API access in the vendor's docs before planning an integration into an existing development workflow.

For a prototype, the browser interface keeps prompting, code generation and hosting in one place. Bolt says it can test, refactor and iterate automatically, and its built-in context management is intended to handle larger projects. The homepage describes projects as 1,000 times larger than before, without defining a project-size unit. This is not the same as a documented build pipeline: teams should check how they can inspect, version, export and deploy generated code before making Bolt a required step. GitHub is a listed starting point, but confirm the exact supported workflow in the vendor's documentation.

The Free plan costs $0 per month and includes a 300K daily token limit, 1M tokens per month, 10MB file uploads, up to 333k web requests and Bolt branding on websites. It is suitable for evaluation, but those token and request limits matter if you build regularly.

Pro costs $25 per month. It starts with 10M tokens per month, has no daily token limit, allows 100MB uploads and up to 1M web requests, supports custom domains and rolls unused tokens into the next month. Teams costs $30 per user per month and adds centralized billing, team access management, private NPM registries and design-system knowledge with per-package prompts. Enterprise pricing is custom and includes SSO, audit logs, compliance support and priority support.

The pricing page says annual billing saves up to 28%, but does not list exact annual prices. It also does not list token overage rates, so confirm the cost of additional usage before relying on Bolt for production workloads. Teams costs rise with each member.

Bolt.new is presented as a hosted service with Bolt Cloud infrastructure. The vendor doesn't publish its open-source licence, self-hosting option or data-export terms on the homepage. Do not assume you can run Bolt's builder or backend in your own environment.

The homepage lists GitHub as a way to start a project, which may fit a code-centric workflow, but it does not establish what can be exported or how the connection works. Before committing, ask the vendor how to retrieve generated code, data, uploaded files and database contents, and whether you can deploy the project outside Bolt Cloud. Generated code may give you a route to continue elsewhere, but backend services and hosting can create migration work. The less your project depends on Bolt-specific services, the easier it is likely to move.

The Free plan's monthly token and web-request limits, along with file-upload caps, set concrete boundaries for development. Pro raises those limits, but the vendor does not list token overage rates on the pricing page. Check how usage is measured and what happens at each limit before putting a customer-facing project on the service. Enterprise lists SSO, audit logs and compliance support; those requirements are not listed on the other plans.

Bolt claims automatic testing and refactoring cut errors by 98%, and says its context management supports projects 1,000 times larger than before. Those are vendor claims, not service-level reliability guarantees. Reddit users report empty projects, bugs, wasted tokens and difficulty as projects grow. Other commenters say careful prompting and managing project context can help, but those are individual reports, not a support commitment. Review generated code, test deployments and verify database export and recovery paths before depending on Bolt in production.