Hangzhou-based K2Lab has closed a tens-of-millions-yuan Angel+ round led by Kaite Capital, its third financing in six months, as the company’s commission-only AI agent for TikTok creators gains commercial traction.
The round was announced in late October 2026. Previous backers include Yunshi Capital, which led the seed round in January, and Huakong Fund, which led the angel round in April. The fresh capital will go toward overseas expansion, upgrading model algorithms and engineering capabilities, building a merchant agent and merchant alliance ecosystem, and developing an AI-native organization.
What K2Lab Does
K2Lab was founded in 2025 by Wang Ming, a former vice president of DingTalk and the youngest person to hold that title at the Alibaba-owned platform. Its flagship product, Moras, is a personal commerce agent for creators on TikTok. It handles product selection, script generation, video creation, publishing, and performance analysis.
Moras’s pricing model is unusual: creators pay nothing upfront. There is no subscription fee and no token charge. K2Lab earns a commission only when a sale is completed, splitting revenue with the creator after the merchant pays. The company has positioned this as a “results-first” alternative to the subscription and usage-based models common among AI applications.
Traction and Economics
K2Lab has disclosed several operating metrics. In its first public-testing quarter, Moras facilitated over $10 million in gross merchandise value (GMV) for creators on TikTok, with nearly 100,000 creators registered. The company says its retained commission revenue covered computing costs during the testing phase—a milestone that many AI agent startups have struggled to reach.
Moras became a TikTok Gold Service Provider within three months and is preparing for the Black Friday shopping season with partnerships covering more than 1,100 head merchants across eight industries.
Technical Approach
K2Lab has built a proprietary multimodal content model called K2 Prism, designed to parse reference videos into structured tags product display, video style, opening hooks, proof elements, and closing calls-to-action. The company describes its technical route as “model-application synergy,” with in-house capabilities in understanding, prediction, and scheduling, while using external models for some content-generation steps.
K2Lab has also brought in Zhao Qiang, a former Meituan technical executive, to oversee model, algorithm, and engineering work, joining founding CTO Zhao Xianlie.
The Financing Strategy
K2Lab’s funding cadence three rounds in six months, each led by a different institutional investor reflects a deliberate approach. Wang Ming has said that AI-native teams should prioritize financing efficiency rather than waiting for a single large round, using rapid raises to secure runway while racing to establish a market position. For founders navigating a cautious funding environment, K2Lab’s playbook offers a concrete case study: prove that users are making money, keep the raise small and fast, and let commercial evidence drive investor interest.
The next test is whether Moras can scale its commission-based model without proportional increases in compute costs. K2Lab has not disclosed revenue figures beyond the claim that commissions covered compute costs during testing. Whether “results-first” AI agents become a durable model or remain a niche experiment will depend on that unit-economics question.
