← 返回公开榜单
ac

accomplish-ai/accomplish

Accomplish™ (formerly Openwork) is the open source Al coworker that lives on your desktop

本周新增
36 星
近 3 周倍数
0.1x
距上次推送
237 天
距上次发布
264 天
许可证
MIT

星标增长

每根柱子是这个周期记录下来的 star 增长:优先取逐个 stargazer 到达时间分桶的到达数,还没被 stargazer 扫描覆盖的仓库改用该周期的 star 净增;两个写者都没记录的周期画成虚线空档,而不是当成 0。柱高按各自窗口的峰值归一,两张图不共用坐标轴。指向任意一根柱子, tooltip 里是该周期的九项指标:当前值与周期内的净增。

指向或点击一根柱子,看当天的九项指标。

2/232/263/13/43/73/103/133/16

2026 年 2 月 25 日 新增 284 个星标; 2026 年 2 月 27 日 新增 155 个星标; 2026 年 2 月 28 日 新增 26 个星标; 2026 年 3 月 1 日 新增 22 个星标; 2026 年 3 月 2 日 新增 292 个星标; 2026 年 3 月 3 日 新增 143 个星标; 2026 年 3 月 4 日 新增 80 个星标; 2026 年 3 月 5 日 新增 38 个星标; 2026 年 3 月 8 日 新增 82 个星标; 2026 年 3 月 10 日 新增 40 个星标; 2026 年 3 月 11 日 新增 8 个星标; 2026 年 3 月 16 日 新增 36 个星标

窗口内最高的一天是 2026 年 3 月 2 日,新增 292 个星标。

证据时间轴

这个项目还没有记录到异动、发布或许可证变化。

README
The document stored for this repository, with its Chinese translation when one has been produced

English | 中文 | 日本語 | 한국어 | Русский | Español | Türkçe | العربية | Bahasa Indonesia | தமிழ் | हिन्दी

Accomplish - Open source AI desktop agent that automates file management, document creation, and browser tasks with your own AI API keys

MIT License GitHub Stars GitHub Issues Last Commit Download for macOS (Apple Silicon) Download for macOS (Intel) Download for Windows 11 Discord

Accomplish™ - Open Source AI Desktop Agent

Accomplish is an open source AI desktop agent that automates file management, document creation, and browser tasks locally on your machine. Bring your own API keys (OpenAI, Anthropic, Google, xAI) or run local models via Ollama.

Runs locally on your machine. Bring your own API keys or local models. MIT licensed.

Download for Mac (Apple Silicon) · Download for Mac (Intel) · Download for Windows 11 · Accomplish website · Accomplish blog · Accomplish releases




What makes it different

🖥️ It runs locally

  • Your files stay on your machine
  • You decide which folders it can touch
  • Nothing gets sent to Accomplish (or anyone else)

🔑 You bring your own AI

  • Use your own API key (OpenAI, Anthropic, etc.)
  • Or run with Ollama (no API key needed)
  • No subscription, no upsell
  • It's a tool—not a service

📖 It's open source

  • Every line of code is on GitHub
  • MIT licensed
  • Change it, fork it, break it, fix it

⚡ It acts, not just chats

  • File management
  • Document creation
  • Custom automations
  • Skill learning



What it actually does

📁 File Management✍️ Document Writing🔗 Tool Connections
Sort, rename, and move files based on content or rules you give itPrompt it to write, summarize, or rewrite documentsWorks with Notion, Google Drive, Dropbox, and more (through local APIs)
⚙️ Custom Skills🛡️ Full Control
Define repeatable workflows, save them as skillsYou approve every action. You can see logs. You can stop it anytime.

Use cases

  • Clean up messy folders by project, file type, or date
  • Draft, summarize, and rewrite docs, reports, and meeting notes
  • Automate browser workflows like research and form entry
  • Generate weekly updates from files and notes
  • Prepare meeting materials from docs and calendars

Supported models and providers

  • Anthropic (Claude)
  • OpenAI (GPT)
  • Google AI (Gemini)
  • xAI (Grok)
  • DeepSeek
  • Moonshot AI (Kimi)
  • Z.AI (GLM)
  • MiniMax
  • Amazon Bedrock
  • Azure Foundry
  • OpenRouter
  • LiteLLM
  • Ollama (local models)
  • LM Studio (local models)

Privacy and local-first

Accomplish runs locally on your machine. Your files stay on your device, and you choose which folders it can access.


System requirements

  • macOS (Apple Silicon)
  • Windows 11



How to use it

Takes 2 minutes to set up.

StepActionDetails
1Install the AppDownload the DMG and drag it into Applications
2Connect Your AIUse your own Google, OpenAI, Anthropic (or other) API key — or sign in with ChatGPT (Plus/Pro). No subscriptions.
3Give It AccessChoose which folders it can see. You stay in control.
4Start WorkingAsk it to summarize a doc, clean a folder, or create a report. You approve everything.





Screenshots and Demo

A quick look at Accomplish on macOS, plus a short demo video.

Accomplish demo - AI agent automating file management and browser tasks

Watch the demo →


FAQ

Does Accomplish run locally?
Yes. Accomplish runs locally on your machine and you control which folders it can access.

Do I need an API key?
You can use your own API keys (OpenAI, Anthropic, Google, xAI, etc.) or run local models via Ollama.

Is Accomplish free?
Yes. Accomplish is open source and MIT licensed.

Which platforms are supported? macOS (Apple Silicon) and Windows 11 are available now.




Development

pnpm install
pnpm dev

That's it.

Prerequisites
  • Node.js 20+
  • pnpm 9+
All Commands
CommandDescription
pnpm devRun desktop app in dev mode
pnpm dev:cleanDev mode with clean start
pnpm buildBuild all workspaces
pnpm build:desktopBuild desktop app only
pnpm lintTypeScript checks
pnpm typecheckType validation
pnpm -F @accomplish/desktop test:e2ePlaywright E2E tests
Environment Variables
VariableDescription
CLEAN_START=1Clear all stored data on app start
E2E_SKIP_AUTH=1Skip onboarding flow (for testing)
Architecture
apps/
  desktop/        # Electron app (main + preload + renderer)
packages/
  shared/         # Shared TypeScript types

The desktop app uses Electron with a React UI bundled via Vite. The main process spawns OpenCode CLI using node-pty to execute tasks. API keys are stored securely in the OS keychain.

See CLAUDE.md for detailed architecture documentation.




Contributing

Contributions welcome! Feel free to open a PR.

# Fork → Clone → Branch → Commit → Push → PR
git checkout -b feature/amazing-feature
git commit -m 'Add amazing feature'
git push origin feature/amazing-feature



Accomplish website · Accomplish blog · Accomplish releases · Issues · Twitter


MIT License · Built by Accomplish


Keywords: AI agent, AI desktop agent, desktop automation, file management, document creation, browser automation, local-first, macOS, privacy-first, open source, Electron, computer use, AI assistant, workflow automation, OpenAI, Anthropic, Google, xAI, Claude, GPT-4, Ollama

页面上的每个数字都来自记录下来的 stargazer 到达时间,没有评分公式,也没有 AI 判定。 任何结论都可以用 /api/v1/repos/{id}/stats 里的原始增量自己重算。