HonmaruAI

• 

  Honmaru AI app icon — 1024 x 1024, uncropped.

• 

  Decision feed: review the request, context and recommendation. Native iPhone screenshot; fictional, brand-neutral sample data.

Inspiration

Small teams make important decisions throughout the day, but the request, context, and final answer often live in different conversations. For a small team shipping a launch, a single unanswered budget request can stall the next step. We wanted a focused place where someone can understand what needs a decision and respond without reconstructing the entire discussion. As AI produces more drafts, summaries, and proposals, the bottleneck shifts toward shared context and clear human decisions.

What it does

Honmaru AI brings the familiarity of a card-based feed to team work. Our direction is an AI-native work OS: people and named AI agents share context, while people remain responsible for decisions.

Honmaru AI is a native iPhone app for team requests and approvals. Write a request, review it before sharing, and bring the decision into a focused card-based feed. The recipient can inspect the context and respond. Team and history views help people understand who is involved and revisit requests later.

Version 1.1.0 also brings richer team chat: threads, reactions, pinned messages, Later, and text formatting. Team members can mention a named AI agent for research, drafts, and summaries in the thread. Asking @AI a question produces a reply; it creates a decision card when asked.

Messages, notifications, and Activity can be translated into the selected language. Read state and notifications sync across devices, helping teammates keep track of what still needs attention.

A concrete use case is an onboarding-video reshoot: a marketer requests a budget, the reviewer sees the amount and supporting context, and a human makes the approval decision. A named AI agent can then help draft next steps in a team conversation. The film illustrates these capabilities in separate demo segments; it does not claim an automated end-to-end integration.

The interface keeps the decision prominent and the supporting detail close at hand. Light and dark appearances support the same workflow.

How we built it

The iPhone app uses SwiftUI. Its backend runs on Cloudflare Workers, with Durable Objects for real-time coordination and D1 for persistent data. AI-assisted request processing uses OpenAI. RevenueCat handles the app's subscription purchase and entitlement integration, with monthly and yearly Pro products and a restore-purchases path.

Challenges we ran into

A team approval app needs more than a polished screen. Requests must reach the intended person, account changes must not expose another person's workspace, and a decision must remain consistent after reconnection. We also had to keep the iPhone client and backend compatible as the product evolved.

Accomplishments

We shipped Honmaru AI to the US App Store during Shipaton. Apple's public listing records the initial release on September 21, 2026. The product brings request creation, review, context, and decisions into a native mobile experience. Apple approved version 1.1.0, and we released it on September 29, 2026, adding richer chat, team AI agents, and translation.

What we learned

Making a decision easier starts with making the request clearer. Separating composition, review, and response helped us focus the interface on what the next person needs to know. We also learned to validate purchase and account behavior separately from visual polish.

What's next

We plan to learn from real teams, refine the request-to-decision flow, and improve onboarding and subscription clarity based on observed use.

About the demonstration

The demo uses native iOS screens from the version 1.1.0 release source, running in an iPhone simulator. Decision cards are shown in the app's local demo mode. The chat and AI segments use a separate server-backed demonstration workspace with fictional content and a real AI-agent response. A capture-only startup harness opens these native screens; it does not recreate their UI. Sample source labels in recaptured decision screenshots were made brand-neutral for the film.

RevenueCat recognizes the dedicated judge account's complimentary Pro entitlement. No purchase or restore transaction is shown. Private judge credentials are supplied separately and are not included in this video or public description.

Built With

• cloudflare
• openai
• revenuecat
• swiftui