Rabbitpre Tech (兔展智能) raises Series F, launches SkillsUI to bridge legacy enterprise systems with AI agents
The AMW Read
Novelty is incremental — agent middleware is an established pattern — but significance is segment-level because it demonstrates a viable enterprise deployment pathway that updates the 'AI entry layer' open debate.
Rabbitpre Tech (兔展智能) raises Series F, launches SkillsUI to bridge legacy enterprise systems with AI agents
Shenzhen-based Rabbitpre Tech (兔展智能), a visual AI company serving over 41 million users, has closed an undisclosed Series F round and released SkillsUI, a product that acts as an AI orchestration layer atop existing enterprise IT stacks. SkillsUI connects APIs from legacy ERP, OA, and CRM systems, decomposing their capabilities into callable "Skills" that an AI agent can invoke based on natural language user requests. The company has previously built the "Tuling" (兔灵) multimodal model — the first visual-spatial intelligence model registered in Guangdong Province — and the open-source Open-Sora Plan, which surpassed 26 million downloads.
Why it matters: Rabbitpre Tech exemplifies the emerging "agent middleware" pattern where startups build a unified AI entry layer over entrenched enterprise systems rather than replacing them. This approach directly addresses the bottleneck between conversational AI and action-capable AI — a gap that is becoming the central deployment challenge in enterprise AI. The company's self-developed multimodal foundation model provides the intent-understanding and tool-calling reliability that makes the middleware viable, signaling a convergence of model capability and systems integration that could reshape enterprise software economics.
Grounded expert take: Rabbitpre Tech's playbook combines two structural forces: the capital-cycle pressure to demonstrate enterprise ROI from years of LLM investment, and the recognition that system-level orchestration — not chat — is the last-mile bottleneck in enterprise AI adoption. By positioning SkillsUI as an API-based layer that preserves existing IT investments, the company avoids the lock-in concerns that hinder many AI platforms. The core insight is that enterprise AI value will flow to whoever owns the entry layer between users and legacy systems, not to those who build yet another chat interface. The 41 million user base and Series F round suggest investors see this bet as viable.