← Sticker Wall AGENT · DETAIL

Wrtn Korea AI Portal and Agent Store Hits 30 Million RMB Monthly Revenue

Workflow: Users initiate job hunting, learning, or creative tasks daily via Wrtn's primary interface using natural language. The p

AGENT

Key Fields

FIELD STAMPS

IndustryEducation / Knowledge

RegionGlobal（韩国）

ScaleSME

ChannelOnline

🔧 Workflow

Users initiate job hunting, learning, or creative tasks daily via Wrtn's primary interface using natural language. The platform dispatches multiple large model agents to handle specific needs such as resume optimization, knowledge Q&A, story continuation, and image generation, outputting reusable text or image results. Monetization is a hybrid of free quotas and paid subscriptions. The backend dynamically allocates different model agents based on user history, using a points system and subscription walls to manage costs and conversion.

🛠 Setup Requirements

Requires building or leasing multiple large model APIs with a unified gateway for dispatching. The backend implements user points and subscription systems, while the frontend provides a chat-based interface. The tech stack requires Python or Node.js plus a vector database. An MVP can be launched by an individual or a 3-person team in about 3 months. The key challenge lies in unifying the output formats of different models into a natural language primary interface experience, which requires prompt engineering and lightweight fine-tuning.

🧰 Toolchain

🔧 OpenAI GPT-5 API
🔧 Wrtn proprietary dispatch layer
🔧 Paid subscription and points system
🔧 Korean NLP fine-tuning tools
🔧 User behavior tracking and analytics backend

💰 Revenue

① Corporate level—Wrtn's own subscription and pay-per-use revenue: Korean consumers pay via monthly subscriptions or per-use fees. 2025 revenue was approximately 47.1 billion KRW, or about 30 million RMB per month (based on company financial reports). The North American AI narrative platform OOC has monthly revenue exceeding 10 billion KRW (approx. 76.43 million RMB, based on official company disclosures), though its share of the total market is not public. ② Replicator level—Vertical scenario subscriptions: Individuals or small teams copy the same model to charge monthly subscriptions to single-scenario clients, earning 10,000 to 30,000 RMB per month (based on case studies, not independently verified); share not disclosed. ③ Value-added ecosystem—AI story product pay-per-use: Users pay per use; similar products have an annualized revenue of about $70 million with paid retention over 70% (public company data); share unknown. ④ Opportunities: Seat-based templates and plugins for small local Korean teams, charged by subscription; seat pricing and market share are not disclosed.

💸 Cost

Large model APIs are billed by token, with GPT-5 level interfaces costing approximately $300 to $1,000 per month for base usage. Server and database costs are about $100 per month; the primary cost comes from model calls. If image generation or long-text continuation capabilities are added, API costs may double, requiring the use of free quota caps and caching mechanisms for control.

⏱ Time Investment

After product launch, 1 to 2 hours per day are needed to monitor agent dispatching and user feedback, with half a day per week for iterating prompts and adding new agent templates. Individual developers can operate this part-time, delegating customer service and content moderation to automated rules initially.

🚀 Getting Started

Select a high-frequency consumer scenario, such as Korean resume polishing or short story generation, and build a single-entry chat agent using the GPT-5 API. After launch, use free quotas to attract users, then verify willingness to pay via subscriptions or pay-per-use. Once payment is validated, gradually integrate second and third model agents to form a unified dispatch portal.

🔑 Keys to Success

✅ Multi-model agent aggregation mitigates individual model limitations
✅ Hybrid monetization of free quotas plus subscriptions improves retention
✅ Replacing traditional App navigation with a natural language primary interface
✅ Deep cultivation of Korean scenarios to build localized data and user moats
✅ Using high paid retention of story products to validate the feasibility of pay-per-use

⚠️ 风险

⚠️ Large model API costs rise linearly with user volume; gross margin is constrained by upstream pricing
⚠️ Intense local competition in Korea, with giants like Naver also launching AI agent platforms
⚠️ High reliance on a single market; global expansion requires addressing local model compliance and multi-language costs
⚠️ High-frequency calls in pay-per-use models may lead to free users eroding profits

📌 Real Cases

📌 Wrtn Technologies reported 2025 revenue of 47.1 billion KRW, a year-on-year increase of about 14 times, with 100 billion KRW in Series C funding and a valuation of 1.2 trillion KRW (approx. $870 million), serving over 6.5 million Korean users.
📌 Wrtn's AI story product, Crack, sees users spending an average of 2 hours per day with a paid retention rate over 70%, generating approximately $70 million in annualized revenue through a pay-per-use model.
📌 Wrtn Technologies CEO Lee Se-young was elected as the inaugural chairman of the Korea Generative AI Startup Association, which consists of 20 startups aiming to build an AI ecosystem accessible to everyone.

Sources