LobsterAI is an MIT-licensed, Electron-based desktop AI agent from NetEase Youdao that runs tools against your real local files, terminal, and browser and accepts commands from your phone through messaging apps — built for developers, analysts, and office teams who want agent automation to stay on their own machine.

What it is

LobsterAI is an open-source desktop Agent built by NetEase Youdao, written in TypeScript and shipped as an Electron application. It splits into two layers: Cowork is the product and session layer inside the desktop app, and OpenClaw is the runtime and gateway underneath it. That split lets LobsterAI keep local persistence, permissions, UI state, artifacts, agents, memory, and IM bindings in the desktop client while OpenClaw handles agent execution.

The concrete problem it solves is the gap between a chat assistant that describes work and an agent that actually performs it in your working environment. Instead of relaying instructions by hand between an LLM conversation and a local toolchain, LobsterAI operates directly on local files, terminal commands, browser workflows, documents, spreadsheets, slides, IM channels, scheduled jobs, and project workspaces. It replaces that manual relay, and it asks for approval before sensitive actions such as file operations, terminal commands, or network access.

Key capabilities

• Desktop Cowork sessions that run long-form Agent tasks against local projects, stream progress, keep session history, render tool output, and gate sensitive actions behind approvals.
• 28 built-in skills configured in SKILLs/skills.config.json, including web search, Word documents, spreadsheets, PowerPoint, PDF processing, Remotion video generation, browser automation, image/video generation, stock research, content writing, email, weather, and skill creation.
• Multi-Agent workflows with per-agent identity, model choice, skills, working directory, enabled state, and IM bindings, keeping a Main Agent for general work and specialized Agents for repeatable roles.
• Expert Kits that package capability selections and references for common workflows, selectable independently from direct skills so a workflow can combine curated kits with individual tools.
• MCP server connections that store user-configured servers locally and sync enabled servers into OpenClaw.
• Scheduled tasks created either by conversation or through the scheduled task UI, for daily news digests, inbox summaries, website monitoring, weekly reports, and other repeatable work.
• IM remote control from WeChat, WeCom, DingTalk, Feishu/Lark, QQ, Telegram, Discord, NetEase IM, NetEase Bee, POPO, and email, plus rich artifact preview for HTML, SVG, images, video, Mermaid diagrams, code, Markdown, text, documents, and local service artifacts.

Who uses it and how

• Individual analysts working from local spreadsheets point it at a file such as product-growth.xlsx to build a visual dashboard and summarize growth drivers.
• Operations staff automate recurring browser checks, opening an ads dashboard every morning to review spend and conversion anomalies and summarize likely causes.
• Recruiters and hiring teams turn a folder of resumes into a screening sheet and shortlist candidates against a job description.
• Teams standardized on Chinese IM stacks bind different accounts or channels to different Agents and drive those Agents from a phone over WeChat, WeCom, DingTalk, or Feishu/Lark.
• Builders who need small local tools generate systems, such as a local inventory system that records purchases and sales, calculates stock and profit, and opens in the browser.
• Anyone with repeatable reporting needs schedules work conversationally, for example collecting yesterday's AI news every weekday at 9 AM and sending a concise digest.

Getting started

The README points to downloading LobsterAI from the project homepage at https://lobsterai.youdao.com, which is the documented way to obtain the desktop application. From there the app starts the OpenClaw runtime and syncs enabled skills, MCP servers, and IM gateways itself.

How it compares

LobsterAI builds on OpenClaw rather than competing with it: OpenClaw is the execution runtime and gateway, and LobsterAI is the desktop product and session layer on top. The provided facts name no other comparable desktop-grade agent for positioning, so on this registry it stands on its own.

When to use it — and when not to

A self-hoster must operate an Electron desktop application with local SQLite persistence and the OpenClaw runtime the app starts and repairs, so this is not a headless server or CI-oriented tool, and teams wanting a managed hosted service or browser-only access should look elsewhere. The project carries 601 open issues against 6,047 stars and 968 forks, which indicates an active but still rough surface with a large bug backlog. The README excerpt is also thin on installation specifics beyond the homepage download, so anyone evaluating it should expect to learn the setup from the app and the repository rather than from a detailed operations guide.

# 1. Build (applies the pinned patches, prunes to ~160 MB)
npm run dsh:runtime:mac-arm64

# 2. Pack; prints the sha256 and size
npm run dsh:runtime:pack mac-arm64

# 3. Upload vendor/dsh-dist/dsh-runtime-<version>-mac-arm64.tar.gz to the CDN,
#    then record where it landed. Digest and size come from the local manifest,
#    so they cannot drift from the bytes that were packed.
npm run dsh:runtime:url mac-arm64 "https://cdn.example.com/<uploaded>"

# 4. Confirm the URL serves exactly those bytes
npm run dsh:runtime:verify-urls mac-arm64

# Full installer for a distribution channel
npm run dist:win:channel -- --keyfrom <channel> [--silent]

# Web installer: a small stub that downloads the package from your CDN
npm run dist:win:web -- --keyfrom <channel> [--silent] [--pkg-base-url <cdn-dir> | --pkg-url <package-url>]