What Is LobsterAI? A Complete Guide to NetEase's All-Scenario Office Assistant Agent and Open-Source Desktop-Level Agent

What Is LobsterAI?

LobsterAI is NetEase's all-scenario office assistant Agent and the first open-source desktop-level Agent from a major Chinese tech company. Its key principles are 'local-first' and 'secure and trustworthy': user conversations, configurations, and memory are primarily kept on the local machine, and tool call processes are controllable and auditable.

Unlike chat assistants that only answer questions, LobsterAI works in real work environments. It can connect to local files, terminals, browsers, and project environments to handle document processing, data analysis, web page generation, PPT creation, browser automation, scheduled tasks, and other scenarios.

Product Positioning: Desktop-Level Agent Office Assistant

LobsterAI positions itself as a 'desktop-level Agent office assistant.' Desktop-level means it is not limited to a web dialog box; it can connect to files, terminals, browsers, and local projects to complete complex office tasks directly in the user's computer environment.

'Local-first' is the core of this positioning. User conversations, configurations, and memory are primarily kept on the local machine. During execution, tool call records can be reviewed to make it easy to confirm what was done at each step.

Core Features

Automatic Task Breakdown and Execution

After users describe their needs in natural language, LobsterAI automatically breaks down the steps and completes them one by one. Tasks such as building an inventory management system, creating a data dashboard, and organizing a resume screening sheet can be handed to it to complete step by step.

Document and Content Creation

In document and content scenarios, it supports content summarization, report writing, PPT creation, copy generation, and material organization. For roles that require repeatedly reading materials and consolidating information, these capabilities can directly contribute to daily output.

Data Processing and Web Page Generation

LobsterAI can read local files such as Excel and PDF, then perform data analysis, create visualization pages, organize tables, and extract key information. In other words, raw spreadsheets and documents go in, and finished pages and conclusions come out.

Browser Automation

It can open web backends, check data, identify anomalies, and summarize conclusions. Repetitive tasks such as operations work, ad campaign management, and data monitoring are well suited to this approach for reducing manual work.

Scheduled Tasks

LobsterAI supports setting up recurring tasks, such as compiling AI tech news every day or checking ad backend data on a schedule. Handing fixed-cadence work to scheduled tasks reduces the burden of having to remember to do things at certain times.

Multiple Agents and Sub-Agents

It can create specialized agents for stock assistance, content creation, lesson preparation and exam generation, and more, then use sub-agents to collaborate on complex tasks. A main task is split among multiple specialized roles that work together to complete it.

Expert Suites

Expert suites install capability packs according to work scenarios. They combine related skills, MCP services, or connectors and can be enabled directly in conversations. For example, a built-in computer operation suite can be used to control local desktop applications.

Skill System and MCP Support

The skill system provides fine-grained workflow capabilities and supports custom skills to extend specific tasks. MCP support connects external tools and data sources through MCP services, expanding the range of tools the Agent can call.

Remote Task Delegation to Your Computer

LobsterAI supports remote task execution on your computer through IM tools such as WeChat, WeCom, DingTalk, and Feishu. Even when you are not in front of your computer, you can still assign tasks to it.

How to Use LobsterAI

1. Download and install: Visit the official LobsterAI website at https://lobsterai.youdao.com/ and download the client. Official versions are available for MacOS with Apple silicon, MacOS with Intel chips, and Windows.
2. Chat directly: After installation, open the client and describe the task or ask a question directly in the input box. Examples include 'Help me organize the resumes in this folder,' 'Make a visualization page based on this Excel file,' and 'Compile AI news every morning at 9:00.'
3. Grant permissions: Authorize LobsterAI to access local files, browsers, terminals, or related tools as needed for the task. During execution, you can view tool call records to keep operations under control.
4. Advanced configuration: Use features such as scheduled tasks, expert suites, skills, MCP, and My Agent to build an automation assistant that better fits your workflow.

Target Users and Typical Scenarios

• Office workers: Users who frequently handle documents, spreadsheets, PPTs, material summaries, and cross-tool collaboration.
• Operations and marketing professionals: Tasks such as ad backend checks, data organization, anomaly analysis, competitor research, and daily or weekly reports.
• Content creators: Topic research, copywriting, content summarization, script organization, and material planning.
• Developers and technical staff: Local project handling, web page generation, automation scripts, terminal operations, and exploring open-source Agent capabilities.
• Small and medium-sized businesses and individual operators: Building lightweight systems, organizing inventory and sales data, creating local tools, and increasing automation of repetitive work.

Strengths

• Local-first, with more control over data: Conversations, configurations, and memory primarily stay on the local machine, and tool call processes are auditable.
• Can enter real work environments: Connects to files, terminals, browsers, and local projects instead of staying only at the conversation layer.
• Covers the full range of office scenarios: Documents, data, web pages, PPTs, browser automation, scheduled tasks, and more are supported in one workflow.
• Extensible: Skills, expert suites, MCP, and multiple agents and sub-agents provide extensibility at different levels.
• Open-source and remote task delegation: As the first open-source desktop-level Agent from a major Chinese tech company, it also supports assigning tasks remotely through IM tools such as WeChat, WeCom, DingTalk, and Feishu.

Limitations

• Depends on permission grants: Entering real work environments requires authorizing local files, browsers, terminals, or related tools per task. Permissions are a prerequisite for use.
• Capabilities depend on configuration: The range of callable tools depends on how skills, expert suites, and MCP are configured. These need to be installed and adjusted per scenario.
• Changeable information should be verified independently: Details such as plans, paid pricing, and platform versions may change. Please refer to official channels for real-time information.

Pricing Verification Reminder

This page does not provide specific pricing figures. For complete plans and paid pricing, please refer to the official website's 'Pricing' page or what is displayed in the client. Before purchasing or subscribing, please check the latest information at https://lobsterai.youdao.com/.

Frequently Asked Questions

Which systems does LobsterAI support?

The official website provides client versions for MacOS with Apple silicon, MacOS with Intel chips, and Windows. You can download and install them from https://lobsterai.youdao.com/.

Where is LobsterAI data stored?

It emphasizes a local-first approach. User conversations, configurations, and memory are primarily kept on the local machine, and tool call processes are controllable and auditable.

Can I use WeChat or DingTalk to assign tasks remotely?

Yes. LobsterAI supports remote task execution on your computer through IM tools such as WeChat, WeCom, DingTalk, and Feishu.

Can it read local Excel and PDF files?

Yes, it can read local files such as Excel and PDF, then perform data analysis, create visualization pages, organize tables, and extract key information.

Where can I find pricing?

For complete plans and paid pricing, please refer to the official website's 'Pricing' page or what is displayed in the client.

Alternatives for Reference

If you are comparing similar products, you can consider two directions: one is AI Agent products that claim to be free and have deep research and operation capabilities; the other is the autonomous AI Agent framework open-sourced by Nous Research. The former leans toward out-of-the-box use, while the latter leans toward autonomous frameworks and further development.

When making a choice, first clarify your needs: whether local-first is required, whether browser and terminal permissions are needed, whether MCP should be used to connect external data sources, and whether scheduled tasks and multi-agent collaboration are required. List these criteria and compare them with actual demonstrations to make a more reliable judgment.

Official Entry

• Official LobsterAI website: https://lobsterai.youdao.com/#/index?keyfrom=aigjj