Capacity.so

What is Capacity.so?

Capacity.so is an AI application generation platform designed for non-technical entrepreneurs, product managers, and developers. Users can describe their ideas in natural language, and the platform will generate the front-end, back-end, database, and authentication components, enabling the deployment of the project as a functional application.

Capacity 3.0 sets full-stack projects as the default configuration, providing each project with its own backend and database. This is different from the earlier approach, which relied heavily on Supabase; now, Supabase is an optional integration to be used when capabilities such as PostgreSQL or real-time subscriptions are needed.

A one-sentence summary

Capacity.so can transform a set of product requirements into a full-stack Web or native mobile application whose code can be viewed, deployed, and exported; it also supports continuous iteration through AI-driven suggestions, the Vibe mode, and the Spec mode.

Development team and product evolution

Capacity was founded at the end of 2024 by its co-founders, Samuel Rondot and Baptiste Studer; the entity listed in the terms of service as being responsible for the service is RocketSocial SASU. The product initially focused on creating websites using natural language, but it has since expanded to include full-stack applications, native mobile applications, and desktop development environments.

Capacity 3.0, released in 2026, introduces an Express backend, JWT authentication, and a separate SQLite database for each project. Users can export the code, deploy it on their own, or hand it over to a development team for further maintenance.

Main functions

Natural language generation applications

Users can directly describe the product’s objectives, target audience, pages, and style. The system first asks key questions regarding the purpose and user experience, and then generates an executable project rather than just a static design draft.

Integration of the front end, back end, and database

By default, a full-stack project generates a responsive frontend, an Express server, and a separate database. When a user requests the addition of a registration function, the AI can create the necessary interface, API routes, and data tables simultaneously.

AI co-creation and fault repair

The AI Co-Founder is used to clarify the scope of a product, answer questions, explain technical concepts, and address operational issues. Users can describe problems in plain language, such as “the button doesn’t work,” allowing the platform to identify and fix the issue.

Real-time preview, logging, and data viewing

The editor provides real-time previews, backend logs, and database views. These interfaces allow users to examine pages, requests, errors, data tables, and query results.

Code download, GitHub, and version restoration

The project’s source code can be downloaded as a ZIP file, or users can connect to their own GitHub account to manage the repository. Historical versions of the code are saved before and after any AI-driven changes, which makes it possible to revert to earlier versions in case of errors.

One-click publishing and custom domain names

Static sites can be deployed on a global distribution network, while full-stack applications run on separate virtual servers. The deployment process takes care of building the application, configuring the server, handling DNS and SSL settings, as well as performing health checks; premium plans also allow the use of one’s own domain name.

Native iOS and Android apps

Mobile applications are developed using React Native and Expo; they can be stand-alone apps or share the same backend and database as web front-ends. During development, they can be previewed using Expo Go or simulators, while the final version is built in the cloud.

Environment variables and external services

The project allows for the storage of AI model credentials, payment information, email details, database data, and credentials for other third-party services via environment variables. The documentation states that these variables are encrypted statically, remain hidden by default, are isolated per project, and are not included in the source code.

Comparison between Vibe mode and Spec mode

Complete workflow

1. First, clarify the target users of the application, its core issues, the necessary pages, and the criteria for success.
2. Choose between the project types of static site, full-stack Web application, mobile application, or Web plus mobile.
3. Submit the first version of the requirements in natural language, and address questions related to the purposes, users, and experience during the AI co-creation process.
4. For complex products, add briefings on additional items, user workflows, and visual guidelines, and then implement them in tasks using the Spec mode.
5. Check pages, interactions, and responsive layouts through real-time previews, while also viewing backend logs and data tables.
6. Add environment variables for payments, emails, AI, or external interfaces, and use test credentials to verify error handling.
7. Test the functionality, permissions, security, data backup, and mobile compatibility; after fixing any issues, create a version from which recovery is possible.
8. Publish to the default address or connect to a custom domain name, and export the source code as a guarantee for delivery and withdrawal.

Getting started guide

1. Create an account and select the option to create a new project; it is advisable to use the free trial version first to see if the platform meets your needs.
2. Write a description of the requirements that includes the target users, the core processes, the pages involved, the data, and the design style; do not just provide the name of the application.
3. After the first version is created, test the main workflow first, and then request changes to pages, fields, permissions, and error messages one by one.
4. For complex functions, switch to Spec mode by breaking down the task into data structures, backend logic, pages, and acceptance criteria.
5. Before release, use multiple test accounts to check registration, logging in, permissions, data isolation, payment callbacks, and error handling processes.
6. Select the server specifications and publish it; then check its health status, logs, resource usage, and the SSL status of the custom domain name.

Mobile app release process

1. Create a new mobile project, or add a mobile client that shares the backend and database with an existing full-stack Web project.
2. Design the navigation for small screens, the loading status, behavior in case of a lost network connection, and permission requests, and preview them on a real device or emulator.
3. Configure the application name, icon, package identifier, and signing credentials for iOS and Android.
4. Start cloud-based building, download the resulting products, and conduct compatibility testing on actual devices in the testing environment.
5. Prepare privacy labels, permission descriptions, store materials, and test accounts, then submit them to the respective app stores.

Which users are it suitable for

• Non-technical entrepreneurs: Use natural language to validate the minimum viable version of SaaS, market platforms, or booking-based products.
• Product Manager: Convert product presentations and user workflows into interactive applications that can be actually tested.
• Small businesses: Develop CRM systems, inventory management tools, reporting functions, forms, and internal approval processes.
• Freelancers and agencies: Quickly create customer websites, demonstration prototypes, and responsive front-ends.
• Mobile product team: Uses React Native and a shared backend to develop iOS and Android clients.
• Creators with development experience: Use AI to generate basic code, which can then be exported to local editors, GitHub, or custom deployment processes.

Typical use cases

• A SaaS prototype with accounts, payment options, a data dashboard, and backend management.
• A small market platform consisting of a product list, a shopping cart, and a checkout process.
• Internal tools for customers, inventory, work orders, forms, appointments, and reports.
• Marketing landing pages, portfolios, event pages, and websites with custom domain names.
• Mobile apps for fitness, tasks, food delivery, reservations, messages, and membership services.
• Products that require shared data between the web management interface and the mobile companion app.

Product advantages

• It extends from the frontend to the backend, databases, and deployment, and is not limited to generating static pages.
• AI will ask about the product objectives and desired user experience before starting development, and the Spec mode allows specifications to be confirmed first before implementation.
• Full-stack projects have their own backend and database, allowing direct access to logs, metrics, and data.
• It supports Web, iOS, and Android; API and data can be shared across full-stack and mobile projects.
• It offers code downloads, GitHub integration, and the option for self-deployment, thereby reducing the risk of platform lock-in.
• The default hosting includes automatic SSL, health checks, real-time logging, and resource metrics.

Usage restrictions and precautions

• AI-generated authentication mechanisms, permission controls, payment processing, database queries, and API logic may contain vulnerabilities; they must be manually reviewed before being deployed in production.
• Each AI message consumes a different number of points depending on the complexity of the task, making it difficult to accurately estimate the total cost of large-scale projects in advance.
• Full-stack servers that exceed the free tier will consume hosting credits based on usage over time; when there are multiple projects, it is necessary to calculate the operational costs separately from the generation costs.
• Free servers have the feature of automatically shutting down when idle; a restart may be required for the first request, so they are not suitable for applications that require extremely low latency.
• Custom domain names are not available in the free tier, and publishing a mobile app requires a separate Apple or Google developer account.
• The fact that the code can be exported does not mean that the maintenance costs disappear; after leaving the platform, one still has to handle building, keys, data migration, deployment, and security updates on their own.
• The platform interface and documentation are primarily in English; the understanding of Chinese prompts, fonts, localization, and content quality should first be verified using examples.

Prices and packages

As of August 21, 2026, the Web platform of Capacity.so relies on a monthly subscription along with one-time credit packs. Each message sent to the AI agent consumes credits based on its complexity, and the purchased credits do not expire.

One-time point package

The official update log states that downloading the ZIP source code is available to all plans, yet the current pricing page lists “full code export” as a benefit associated with the Growth plan. This is likely the difference between ZIP downloads and exports via GitHub; it is necessary to refer to the details of the benefits available in one’s account before making a purchase.

Managed server specifications

For servers of larger capacity, points are deducted on an hourly basis based on their uptime. When the points are exhausted, the application does not stop working immediately; instead, it is automatically downgraded to the 256 MB free version.

Supported platforms

APIs, SDKs, and open-source status

The full-stack applications generated by Capacity can include real API routes, middleware, and authentication mechanisms; mobile and web clients can both use these interfaces. This represents the backend capabilities of a user’s project, and it does not mean that the Capacity platform offers public APIs for third-party developers.

As of the time of verification, no public API references for the Capacity platform, no official SDKs, and no source code repositories for its core products were found. The platform itself should be regarded as a closed-source commercial service; exporting users’ projects to GitHub does not mean that the Capacity product is open source.

Data security and privacy

The environment variable documentation states that keys are stored in an encrypted format, a default mask is used, isolation by project is applied, and data is transmitted over encrypted connections. Users should still use dedicated credentials with limited permissions, update them regularly, and avoid exposing them in prompts, screenshots, or client code.

The official privacy policy states that the content created by users belongs to those users, and that no personal data is collected except for the information provided by the users themselves. However, the more recent version of the terms of service specifies that such data can be used to train and improve AI systems as well as to support service enhancements; there are discrepancies between the statements in these two documents that require further clarification.

Safety checklist for production projects

• Real customer, medical, financial, or identity data is not used in the initial prototype.
• Create revocable, low-privilege keys that are isolated by environment for each external service.
• Check the authentication, resource ownership, input validation, and rate limiting for each API route.
• Establish processes for backing up, migrating, deleting, exporting, and recovering from disasters for databases.
• Assess the aspects of AI training, content processing, sub-processors, data location, and deletion procedures outlined in the terms.
• Complete dependency scanning, permission testing, log masking, and professional code review before going live.

Basic information

Recommendation score

Its rating is 4.3 out of 5 points. Capacity.so has evolved from an AI-based website builder into a full-stack platform that includes a separate backend, database, mobile applications, and production hosting; the code can be exported, which provides convenience for handover in the future.

The main uncertainty lies in the fact that it is difficult to estimate precisely in advance the computational resources required for complex tasks; moreover, the production code generated by AI still needs to undergo professional testing and security reviews. Teams dealing with sensitive code and data should also request clarification from the relevant parties regarding the differences between the terms and conditions and the privacy policies before adopting such code.

Frequently Asked Questions

Is Capacity.so free?

It is possible to start using the service without having a credit card; the official website states that the first application can be created for free. For ongoing use, as well as for more advanced hosting options, a subscription or the purchase of credits is required.

Do the points expire?

According to the official price information, points do not expire. However, the rules regarding the renewal and cancellation of monthly plans, as well as point accumulation, should still be based on the information provided on the purchase page and in the account details.

How many points are consumed by one prompt?

There is no fixed price per message. Simple modifications require fewer points, while new features, debugging, and complex operations involve the use of more points.

Do the generated applications have a real backend?

Yes. The default full-stack project for Capacity 3.0 includes an Express server, API routing, authentication, and a separate database – it is not just a frontend demonstration.

Can it be connected to Supabase?

It is possible, but in the current architecture it is an optional integration; it needs to be taken into consideration only when PostgreSQL, real-time subscriptions, or related identity services are required.

Is it possible to export the source code?

It is possible to download the project’s source code, and to connect it to GitHub as well as to one’s own deployment environment. The options related to ZIP downloads, GitHub exports, and custom domain names vary depending on the package chosen.

Can iOS and Android apps be released?

It is possible to create React Native mobile applications, build them in the cloud, and submit them to the two major app stores. Users need to provide their own developer account, signing credentials, as well as the necessary materials for the stores, and these must pass the platform’s approval process.

Is programming knowledge required?

It is possible to create prototypes and common applications without writing any code. However, when real payment processes, sensitive data, complex permission settings, or high availability requirements are involved, professional development skills and security expertise are still necessary.

Are public APIs or SDKs available?

No public platform APIs or official SDKs intended for third-party developers have been found; the APIs generated within a project belong to the user’s own application.

Is Capacity.so open source?

The platform itself is not an open-source project; no official repository containing the source code of its core components has been found. Users can own and export the application code they create, but this is something different from the platform being open-source.

Can it be used for commercial purposes?

The terms of service permit use for both personal and commercial purposes, within the scope of the packages and terms specified. The platform does not claim ownership over the user-generated websites, business logic, or custom implementations. Users remain responsible for the materials they upload, as well as for any third-party code, data, and compliance with relevant regulations.

Are sensitive items suitable for direct upload?

It is not recommended to upload sensitive code or real personal data before completing the contract and undergoing security and privacy reviews. In particular, when there are differences in the provisions regarding data usage between the terms and the privacy policy, companies should seek written clarification first.

Summary

Capacity.so is suitable for individuals and small teams who want to create functional products quickly using natural language. Its feature lies in the fact that it integrates the front end, back end, database, mobile applications, hosting, and code export into a single AI-driven workflow.

When putting it into use, one should not consider only the generation speed; it is also necessary to evaluate factors such as integration overhead, code maintainability, security aspects related to permissions, allocated resources, and the ability to withdraw data. For projects that involve payment processing, personal information, or high availability, professional testing and code review remain essential.