Expert kits. A complete set of skills.
Install a set of skills for product, marketing or data work. Choose the right kit for your task, from writing specs and reviewing competitors to creating charts.
The first open-source desktop-grade Agent among major Chinese tech companies, built by NetEase Youdao and publicly praised by OpenClaw's founder
100% open-source code, transparent capabilities, and auditable permissions, data and execution flows
Built on the open-source Agent framework, with ongoing access to Skills, tools and plugins
Use WeChat, WeCom, DingTalk, Feishu and other IM apps to remotely command LobsterAI 24/7
Connect files, terminal, browser and local projects, then work directly inside your real environment
Sessions, configuration and memory stay on your device, with gated records for tool calls

This is a live, hands-on demo — click anything to try it
Documents, data, presentations. Tell LobsterAI what you need and let it get to work.
From planning to writing code and running it locally, LobsterAI helps you through every step.
Send a message through your favorite chat app. LobsterAI keeps working on your computer and sends you the finished files.
Install a set of skills for product, marketing or data work. Choose the right kit for your task, from writing specs and reviewing competitors to creating charts.
Inspect the source code. Keep sessions, settings and memory on your computer, and manage permissions for local files and tools.
View the source on GitHubSee how LobsterAI uses your computer to take a task from start to finish.
Quick answers about installation, models, remote use and data security.
LobsterAI (Chinese name: 有道龙虾) is an open-source desktop AI agent from NetEase Youdao, built as an all-scenario office assistant. It runs on your computer and works with local files, the terminal, the browser and your projects to handle documents, data analysis, app development, research reports and scheduled tasks.
LobsterAI is built on OpenClaw, the open-source agent framework, and keeps adding Skills, tools and plugins from the OpenClaw ecosystem. On top of it, LobsterAI provides a desktop client that installs in one step and lets you direct tasks remotely through messaging apps such as WeChat, WeCom, DingTalk and Feishu.
Yes. LobsterAI is released under the MIT License on GitHub (netease-youdao/LobsterAI). Its capabilities are transparent, and permissions, data handling and execution can be audited.
View the GitHub repositorymacOS (Apple Silicon and Intel) and Windows. On the download page you can choose the right installer, find previous releases and check which chip your Mac uses.
Go to downloadsLobsterAI offers official plan models, including leading Chinese models such as DeepSeek, Kimi, GLM and MiniMax, ready to use without applying for your own API key. You can also add custom models under Settings → Models with an API key from providers such as DeepSeek, Anthropic or OpenAI, or run local models through Ollama and handle tasks without an internet connection.
Read the model configuration guideYes. After setting up an IM bot in LobsterAI, you can send tasks to LobsterAI on your computer and receive results through chat apps such as WeChat, WeCom, DingTalk and Feishu, around the clock.
Read the IM bot configuration guideConversations, settings and memory are currently stored on your computer, and cloud sync is planned. API keys for custom models stay on your device and are never uploaded, and tool calls are gated and logged. When you use a cloud model, task content is sent to that model service. For more sensitive data, you can switch to a local model through Ollama or a privately deployed model.
The desktop client is free to download and its code is open source. When you use your own API key, the model provider charges you based on usage. LobsterAI also offers credits and subscription plans; see the pricing page for details.
View pricing
Install it and put your computer to work.