Bolt.new

Converts natural language prompts into executable full-stack web applications by invoking an AI agent that generates React/Next.js frontend code, Node.js backend logic, and database schemas. The agent runs code in-browser via WebContainers to validate syntax and functionality before deployment, iterating on the generated code based on execution feedback. Token consumption scales with project complexity (larger codebases consume more tokens per iteration), and the agent supports design system imports from Figma and GitHub to accelerate UI generation.

Runs generated Node.js code and React applications directly in the browser using WebContainers, a sandboxed JavaScript runtime that emulates a Linux environment. The agent automatically executes generated code to validate syntax, test functionality, and detect errors before user review. WebContainers provide filesystem isolation, process sandboxing, and network restrictions, preventing malicious code from accessing the host system. Test results feed back into the agent's iteration loop to refactor and fix errors.

Provides two interaction modes: Plan Mode (where the agent outlines a development strategy before implementation) and Discussion Mode (where the agent and user iterate on requirements and design before code generation). Plan Mode enables users to review and approve the agent's approach before code is generated, reducing wasted token consumption on incorrect implementations. Discussion Mode optimizes token efficiency by clarifying requirements upfront. The specific differences between modes and their impact on token consumption are undocumented.

Generates React Native mobile applications using Expo framework and integrates with Expo services for building, testing, and deploying iOS and Android apps. The agent generates Expo-compatible code with native module support and can configure Expo build services for over-the-air updates and app store deployment. Mobile app generation follows the same natural language prompt interface as web apps, abstracting platform-specific complexity.

Indexes the project filesystem and codebase to provide context-aware code generation and completion. The agent analyzes existing code structure, imports, dependencies, and patterns to generate code that integrates seamlessly with the existing project. Token consumption scales with project size because the entire codebase is indexed and included in the context window. The indexing mechanism and compression strategy are undocumented.

Provides 'Plan Mode' and 'Discussion Mode' features that enable iterative refinement of applications through conversation. Users can discuss design decisions, ask the agent to plan features before implementation, and refine requirements through dialogue. The agent maintains conversation context and can adjust implementation based on feedback without losing project state.

Stores generated and edited Bolt projects in Bolt Cloud infrastructure, providing persistent storage across browser sessions and device access. Projects are associated with user accounts and can be accessed from any browser. Storage limits are 10MB (free tier) and 100MB (Pro tier). Projects can be shared publicly or privately (private sharing requires Pro tier). No documented export format or data portability mechanism; projects are locked into Bolt's infrastructure.

Provides a 'Plan' mode that allows users to discuss and refine application requirements before code generation begins, and a 'Discussion' mode for iterative refinement after generation. The agent can break down complex requirements, ask clarifying questions, and validate understanding before committing to code generation. This reduces iteration cycles by ensuring requirements are clear before implementation.

The agent automatically detects errors from code execution feedback (console logs, test failures, runtime exceptions) and refactors generated code to fix issues. It uses a feedback loop where execution results inform the next generation step, allowing the agent to retry failed operations, fix syntax errors, and optimize code without user intervention. The claimed '98% error reduction' suggests automated testing and validation, though the specific mechanism (unit tests, linting, type checking) is undocumented.

Imports design systems and UI component libraries (Material UI, Chakra UI, Shadcn UI, Washington Post Design System, Porsche Design System) and Figma designs directly into the code generation pipeline. The agent uses design tokens, component definitions, and visual specifications to generate React components that match the design system's styling and structure. GitHub repository imports allow the agent to extract existing design systems and apply them to new generated code.

Automatically generates code to integrate with external services and APIs including Stripe (payments), Supabase (database and auth), Netlify (hosting), Google SSO (authentication), Expo (mobile), and MCP servers (custom integrations). The agent handles API credential management, authentication flow setup, and boilerplate code generation for common integration patterns. Integration configuration is abstracted from the user; the agent infers required API calls and data structures from the application description.

Implements a token-based billing model where each interaction (code generation, iteration, refactoring) consumes tokens from the user's monthly or daily quota. Token consumption scales with project size (larger codebases consume more tokens per message). Free tier provides 300K tokens/day and 1M tokens/month; Pro tier provides 10M tokens/month with no daily limit and rollover to next month; Teams tier allocates per-member token budgets; Enterprise tier offers custom allocations. The agent tracks token usage per project and per interaction, allowing users to monitor consumption and optimize prompts.

Provides team workspaces (Teams plan) where multiple developers can collaborate on shared projects with granular permission controls (admin, editor, viewer roles). Projects can be marked private or public; Pro+ users can share sites privately with specific collaborators. Teams plan includes private NPM registries for dependency management and audit logs for compliance. Real-time collaboration mechanism is undocumented, but projects support concurrent editing and version history tracking.

Tracks version history of projects within Bolt, allowing users to view and revert to previous states. GitHub integration enables users to import repositories, sync changes bidirectionally, and commit generated code to GitHub. The agent can read existing codebases from GitHub and use them as context for generation. Automatic commit behavior and sync frequency are undocumented; manual sync appears to be required for GitHub integration.

Orchestrates deployment of generated applications to Bolt Cloud (managed hosting), Netlify, or user-specified hosting providers. The agent generates deployment configuration (environment variables, build scripts, database migrations) and can trigger deployments automatically after code generation. Bolt Cloud provides managed infrastructure for databases, authentication, and domain management. Netlify integration enables one-click deployment with automatic builds and previews. Deployment mechanism and rollback strategy are undocumented.

Accepts natural language prompts via chatbox interface and supports file uploads (up to 10MB free, 100MB Pro) to provide context for code generation. Users can upload design files, existing codebases, or configuration files to inform the agent's generation decisions. File upload mechanism integrates with the project filesystem, making uploaded files available to the agent for analysis and code generation.