Botflow

Botflow at `botflow.io` is an AI app-building workspace, not the older chatbot product that previously used the same brand and not the unrelated BotFlow services operating on other domains. Its current pitch is straightforward: describe an application in natural language and let an agent write the code, create the backend, start a live preview and deploy the result.¹

The current product spans two development targets. For the web, Botflow generates full-stack React applications with a Convex backend, GitHub integration and one-click deployment to Cloudflare. For mobile, it is building a native iOS path that generates real SwiftUI rather than a JavaScript wrapper; Botflow's own current comparison material labels that native pipeline early access.¹²³

This is a desk review. AiToolMap did not create, debug or publish an application with Botflow and does not claim first-hand measurements of code quality, prompt success rate, latency, deployment reliability or App Store success.

Botflow's native strategy is unusual among browser-based AI builders. The company says it generates real SwiftUI and uses a browser-streamed iOS Simulator, with a managed build/signing and App Store publishing path.³

That is attractive for builders who specifically want Apple's native framework rather than React Native or a wrapped web application. The same Botflow workspace can also produce a full-stack web app, so the product is attempting to cover web and native iOS without creating two unrelated workflows.³

The limitation is maturity. Botflow's own July 2026 comparison describes the native platform and publishing flow as early access. It explicitly acknowledges that a competing product such as Rork Max is further along publicly and says Botflow should not be treated as the safer choice if someone needs to publish immediately.³ That candor is useful, but it also means AiToolMap does not present the iOS path as production-proven.

Android is not part of the current Botflow native offering described in the reviewed first-party material. Builders needing cross-platform mobile today should include that limitation in the purchase decision.³

Botflow is not tied to one model. The current home page visibly promotes model choice and names current OpenAI and Anthropic options alongside other providers. Botflow's comparison pages say nine-plus models can be selected, while the pricing page explicitly supports user-provided OpenAI, Anthropic and Fireworks credentials.¹⁴⁷

A notable option is using an existing Claude Pro/Max subscription through Claude authentication, or Codex authentication, rather than consuming Botflow's own managed credits. Botflow positions this as a way to use the underlying coding agent while paying the model provider directly.⁴⁷

This flexibility is economically attractive and reduces one form of platform lock-in. It also makes Botflow difficult to benchmark as a single AI system: output quality can change with model, prompt, project state and routing. AiToolMap therefore does not import a Claude or GPT benchmark and label it a Botflow score.

Botflow's accessible pricing page currently exposes three plans: Free, Pro and Max. Free allows three active projects. Pro allows 20. Max is listed as unlimited. All plans include unlimited collaborators, encrypted secrets, the full IDE and the ability to bring model keys.⁴

Free requires the user to bring a Convex backend and does not include platform-managed model credits. Pro and Max include platform-managed Convex plus model-credit allotments, with Max described as having the larger allowance. Custom domains are not available on Free, and priority support is reserved for Max.⁴

The pricing page says prices are in U.S. dollars, but the version substantively accessible during this audit did not expose the actual current Pro and Max dollar amounts. AiToolMap therefore records the plan structure but does not guess a price from stale snippets, cached pages or third-party claims. This is less satisfactory than a fully transparent public pricing table and should be rechecked in a future update.

The free tier requires no credit card and can build, run and ship projects, making it possible to evaluate the workflow without an immediate subscription.⁴

Botflow repeatedly emphasizes that users own the generated code and can push projects to GitHub. Its product comparisons also say a user can connect an independent Convex account, reducing dependence on a proprietary Botflow-only backend.¹²⁷

Those are meaningful design choices for a young platform. If Botflow changes pricing, pivots or disappears, standard source code in Git and a backend account controlled by the user are more recoverable than a project that exists only inside a closed builder.

This does not mean migration is frictionless. Generated applications can still depend on Convex APIs, Botflow-specific build assumptions or third-party authentication/deployment services. “You own the code” should be understood as a strong portability signal, not a guarantee that another developer can move the whole application with zero work.

Botflow's privacy policy is effective May 2, 2026 and identifies the operator as Botflow LLC in North Carolina. It says account authentication is handled by Clerk; project files, code, configuration and metadata are stored in a PostgreSQL database hosted on Neon; uploads are handled through UploadThing; and GitHub OAuth tokens are stored to perform requested Git operations.⁸

AI features can send relevant code or conversation content to OpenAI, Anthropic or Fireworks AI depending on the model used. Botflow explicitly directs users to those providers' own privacy policies for provider-side handling.⁸

The policy says Botflow does not sell personal information and does not use project code to train AI models without explicit consent. Account and project data are retained while the account remains active; after account deletion, Botflow says it will delete or anonymize personal data within 90 days except where legal or compliance retention applies.⁸

For security, Botflow says it uses TLS for data in transit and access controls for data at rest. It does not, in the material reviewed here, advertise the extensive certification portfolio typical of mature enterprise vendors. Buyers handling regulated or highly sensitive source code should therefore evaluate the underlying provider terms and their own compliance requirements rather than inferring enterprise certification from general security language.⁸

There is also an infrastructure distinction worth preserving. The privacy policy describes Botflow's own platform storage/hosting subprocessors, including Neon, UploadThing and Vercel, while generated applications are marketed with Convex backends and Cloudflare publishing. These are different layers of the stack, not a contradiction.⁸²¹

Botflow publishes extensively about security risks in AI-generated software and argues that typed backend functions, isolated environments, Git history and scoped credentials reduce common failure modes. Its own material says GitHub is connected through scoped OAuth and that secrets are encrypted.⁴¹

The architecture has sensible properties. Keeping database logic behind server functions can reduce some classes of accidental direct-client database exposure, and maintaining real Git history is plainly useful when an AI change breaks a project.²⁷

But AiToolMap found no independent security assessment of the current Botflow product in the fixed panel. Botflow's articles about other tools' vulnerabilities are not evidence that Botflow itself is vulnerability-free. Generated code remains code: authentication logic, authorization, dependency selection, business rules and secrets still need review before a serious application handles sensitive data.

This is currently Botflow's weakest evidence category.

The 50-source panel did not produce a usable exact-product editorial review, product benchmark, verified-user rating population or consumer satisfaction measure. G2 and Capterra therefore contribute no rating input; neither does any other user-review source. Trustpilot has no exact-product population to use even qualitatively.

Beyond the fixed panel, the founder's public profile describes OpenVibeCode/Botflow as a recent project and documents the earlier chatbot version of Botflow as a separate pre-pivot product.⁶ The Botflow company LinkedIn page currently identifies one employee and a very small follower base while announcing a Product Hunt launch.⁵

These details are useful primarily as maturity context. They mean users should expect a startup-stage product: potentially fast iteration and direct founder involvement, but also higher continuity and support concentration risk than a large established development platform.

Botflow is most interesting for independent developers, technical founders and ambitious non-specialists who want the speed of conversational app building but dislike black-box output and backend lock-in.

It is especially worth evaluating if you already pay for Claude or prefer bringing your own model keys, if Convex suits your backend architecture, or if a future native SwiftUI path matters alongside web development.

Teams building regulated, mission-critical or security-sensitive systems should treat Botflow as a development accelerator rather than a substitute for engineering review. The current external evidence base is not strong enough to justify relying on the builder's own production-readiness claims without validation.