LobsterAI
A full-scenario personal assistant Agent that works for you 24/7
NetEase has quietly released Youdao Lobster, a product that feels quite similar to OpenClaw. I had been thinking about trying it out a few days ago, but since I do not have a Mac mini and also do not have an API key for a major LLM, I eventually gave up on testing it.
What OpenClaw Costs
If you want to run OpenClaw on a very low budget, the cheapest setup can stay within about ¥1,500:
硬件：Beelink EQ12 Pro 或旧笔记本（4核+8GB）
系统：Ubuntu 24.04 LTS
模型：通过 Ollama 运行 Qwen2.5 或 Llama 3.2（4-bit 量化）
成本：仅电费（约 ¥20/月）
For a better experience, the budget is more like ¥4,000–5,000:
硬件：Mac Mini M4（16GB 内存）
系统：macOS Sonoma+
模型：本地 Mistral 7B / Qwen 14B
优势：原生集成、语音唤醒、低功耗
So OpenClaw can cost anywhere from a few hundred yuan to several thousand yuan, depending mainly on token usage and the hardware you choose.
Installation
LobsterAI is much easier to get started with. There are no complicated settings: download it, install it, and you are basically ready to go. If you already have Ollama installed locally, LobsterAI can connect directly to the large model running through Ollama.
Download:
LobsterAI Youdao Lobster
For the local model, I used qwen3-vl:8b. It does work, but honestly, it feels very laggy. If you have an API key for another LLM, it is probably worth trying that instead.
The bots it can connect to currently include DingTalk and Feishu. Given that LobsterAI can be tried at zero cost, what else do you need OpenClaw for? It is worth giving it a spin.
Quick Take
The setup is simple, the configuration is simple, and it is especially suitable for Windows users. Since it can be tested at zero cost, there is not much downside to trying it.