Run LobsterAI: Desktop AI Agent for Real Office Work

// Before you start

What you need

A Windows or macOS computer
An AI API key (OpenAI, Anthropic, or compatible provider)
About 500 MB of free disk space
A stable internet connection for downloads and AI calls

Step 1 of 5

Download the installer for your OS

3 min

LobsterAI ships pre-built installers for macOS and Windows. You do not need Node.js or any developer tools — just grab the file and run it. Go to the GitHub Releases page and download the file that matches your system.

Terminal · mac

$ Open your browser and go to: https://github.com/netease-youdao/LobsterAI/releases

$ Download the file that matches your system:

$ macOS (Apple Silicon): LobsterAI-*-arm64.dmg

$ macOS (Intel): LobsterAI-*-x64.dmg

$ Windows: LobsterAI-*-Setup.exe

What you should see

A .dmg or .exe installer file appears in your Downloads folder.

This might happen

On macOS you may see 'App can't be opened because it is from an unidentified developer'.

Open System Settings → Privacy & Security → scroll down and click 'Open Anyway' next to the LobsterAI entry.

Step 2 of 5

Install and launch LobsterAI

2 min

Run the installer you just downloaded. On macOS, open the .dmg and drag LobsterAI to Applications. On Windows, double-click the Setup.exe and follow the prompts. Then open the app from your Applications folder or Start Menu.

Terminal · mac

$ macOS: Open the .dmg file → drag LobsterAI to Applications → open it from Applications.

$ Windows: Double-click LobsterAI-*-Setup.exe → click Next/Install → launch from the Start Menu.

What you should see

The LobsterAI desktop window opens showing a welcome or onboarding screen.

Step 3 of 5

Enter your AI API key

2 min

LobsterAI needs an AI model to do anything useful. It does not come with a built-in model — you supply your own API key. Open Settings inside the app, find the Model or API section, and paste your key. OpenAI (GPT-4o) and Anthropic (Claude) keys both work. You can get a key from platform.openai.com or console.anthropic.com.

Terminal · mac

$ Inside LobsterAI:

$ 1. Click the Settings icon (gear or profile icon, usually bottom-left).

$ 2. Navigate to 'Model' or 'API Key' settings.

$ 3. Paste your API key into the field.

$ 4. Select your preferred model (e.g. gpt-4o or claude-3-5-sonnet).

$ 5. Click Save.

What you should see

The app confirms the key is saved. The chat or agent input area becomes active.

This might happen

The app shows an authentication error or the agent refuses to run.

Double-check that you copied the full API key with no leading or trailing spaces, and that the key has billing enabled on your provider's dashboard.

Step 4 of 5

Run your first agent task

5 min

Type a plain-English task into the main chat input. LobsterAI will plan the steps, show you what it intends to do, and ask for your approval before touching files or running commands. Start with something low-risk like summarizing a document or doing web research.

Terminal · mac

$ In the main chat input, type a task such as:

$ "Search the web for the top 5 AI news stories today and write a short summary."

$ Press Enter and watch the agent plan and execute the steps.

What you should see

The agent shows a step-by-step plan, executes a web search, and returns a formatted summary in the chat. You may be prompted to approve individual actions.

This might happen

The agent gets stuck or returns an error about a missing skill or tool.

Open Settings → Skills and make sure the relevant skill (e.g. Web Search) is enabled. Some skills require additional API keys such as a search provider key.

Step 5 of 5

Try a file-based task with a real document

5 min

LobsterAI can read local files — spreadsheets, PDFs, Word docs. Drag a file into the chat or reference its path in your message. The agent will read it, process it, and produce output like a summary, a chart, or a new document. It will ask for approval before writing anything back to disk.

Terminal · mac

$ Drag any .xlsx, .pdf, or .docx file into the LobsterAI chat window, then type:

$ "Summarize the key points in this document and list any action items."

$ Press Enter.

What you should see

The agent reads the file, displays a summary in the chat, and optionally saves a new output file to your desktop or working directory after you approve.

This might happen

The agent says it cannot access the file or returns a permissions error.

On macOS, go to System Settings → Privacy & Security → Files and Folders and grant LobsterAI access to your Documents or Desktop folder.

// Status

cooked. baked. worked.

A running desktop AI agent that can search the web, read and summarize your local files, and execute multi-step office tasks with your approval at each sensitive action.

// the honest bit

The honest part

LobsterAI is a young open-source project — some skills (Remotion video, image generation, IM integrations) require extra API keys or third-party accounts that are not free. IM remote control via WeChat or DingTalk requires additional setup inside Settings and those platforms' developer portals, which is beyond a simple first run. The app bundles OpenClaw under the hood; if the bundled runtime fails to start you may see a blank agent panel with no clear error message. Scheduled tasks and multi-agent workflows work well once the basics are running, but expect rough edges in a v1 release. API costs are entirely on you — complex tasks like building a dashboard or generating a slide deck can consume significant tokens.