README.en.md
Rama main · réplica de lectura del repositorio original.
<p align="center"><img src="build/icon.png" width="64" alt="OpenWorkBuddy"></p>
<h1 align="center">OpenWorkBuddy</h1>
<p align="center"><b>A local-first AI office workbench that turns your material into files you can open and edit.</b><br>
For people who prepare reports, decks, spreadsheets and documents, and teams that need self-hosting.<br>
Choose your model, inspect execution, and review the delivered files.</p>
<p align="center"><a href="#quick-start"><b>▶ Make a project update deck</b></a> · <a href="https://github.com/CatCatUncle/openworkbuddy/releases/latest">Download</a> · <a href="https://catcatuncle.github.io/openworkbuddy/?utm_source=github&utm_medium=readme&utm_campaign=first_result">Project website</a> · <a href="#docs">Docs</a> · <a href="README.md">中文</a></p>
<p align="center"><a href="https://github.com/CatCatUncle/openworkbuddy/releases/latest"><img src="https://img.shields.io/github/v/release/CatCatUncle/openworkbuddy?style=flat-square" alt="Published release"></a> <a href="LICENSE"><img src="https://img.shields.io/badge/License-PolyForm%20NC-5b5ff7?style=flat-square" alt="PolyForm Noncommercial license"></a> <a href="https://github.com/CatCatUncle/openworkbuddy/stargazers"><img src="https://img.shields.io/github/stars/CatCatUncle/openworkbuddy?style=flat-square&label=Star&color=5b5ff7" alt="Bookmark the project"></a></p>
<p align="center"><a href="https://www.opensourcealternatives.to"><img src="https://www.opensourcealternatives.to/badge-osa.svg" alt="As seen on Open Source Alternatives" width="378" height="28"></a></p>
<p align="center"><sub>Free for personal non-commercial use, schools and non-profits. Company and personal commercial use require a <a href="COMMERCIAL-LICENSE.md">commercial license</a>. To purchase: <a href="mailto:contact@aijentra.com">contact@aijentra.com</a> · WeChat: <b>agi_smith</b>. Model and media services have their own costs.</sub></p>
<p align="center"><img src="docs/images/demo.gif" width="720" alt="OpenWorkBuddy workflow: submit a task, inspect execution and open delivered files"><br><sub>Workflow demonstration. Measure first success with the task below.</sub></p>
---
<a id="quick-start"></a>
<a id="run-it-in-three-minutes"></a>
## Quick start: make a project update deck
**Prerequisites:** a computer that can run the app and a usable model provider, installed local model, or authenticated local CLI. API use requires your own key. This example uses text and simple shapes; it needs no image/video model, messaging channel, or extra connector. Have PowerPoint, WPS or LibreOffice available to open and check the downloaded deck.
1. **Install and launch.** Choose the macOS / Windows package for your architecture from [Releases](https://github.com/CatCatUncle/openworkbuddy/releases/latest): Windows installer `*-win-setup.exe`, portable `*-win-x64-portable.exe` / `*-win-arm64-portable.exe`, or macOS `.dmg`. On Linux, use source below or [Docker deployment](docs/部署.md).
2. **Connect a model.** Register on first launch. In Settings → Models, enter the provider, exact model ID, API URL and key, then test the connection. An installed local CLI can also be selected under Settings → Agent.
3. **Copy the task and open the result.** Select Craft mode and paste the task below. Preview or download `project-update.pptx` from the results panel, open it in presentation software and check all four slides. Then ask: “Turn the next-week plan on slide 4 into a two-column table. Keep the other slides unchanged.”
```text
Make a 4-slide project update deck saved as project-update.pptx.
Audience: project lead. Project: Example website launch.
Slide 1 | Cover: Example website launch — project update.
Slide 2 | Completed this week: Published the product page; fixed the mobile menu.
Slide 3 | Current progress: Adding the user guide.
Slide 4 | Next week: Finish the English version; collect 5 trial-user feedback responses.
Requirements: 16:9, English, clean business style; editable text,
one main point per slide, readable presentation fonts and no overflowing text.
Keep the supplied facts; do not invent metrics. Confirm the file exists
and provide an openable file in the results panel.
```
**Success means:** a nonempty `project-update.pptx` exists on disk and opens in PowerPoint, WPS or LibreOffice; all four slides match the supplied facts, text is editable and nothing overflows; the follow-up updates slide 4 in the same file. A chat reply or successful installation alone does not satisfy this task. Copyable task files: [English](examples/quick-start/task.en.md) / [中文](examples/quick-start/task.md).
**Timing:** start when installation begins and stop when the first file opens and passes review. Record download, startup, model setup and task wait separately. Three minutes is a design target; actual time depends on network, model readiness and execution. No measured first-success rate has been published yet.
<details>
<summary><b>Run from source: Node.js 18+, npm and Git</b></summary>
```bash
git clone https://github.com/CatCatUncle/openworkbuddy.git
cd openworkbuddy
npm ci
npm start
```
After the server starts, open <http://localhost:3800> and continue with steps 2–3 above. Use `npm run app` for desktop development. Source runs keep data in the repository data directory by default; installed apps use `~/OpenWorkBuddy`. See [Install and launch](docs/安装与启动.md).
</details>
### If you get stuck
| Where it fails | Next step |
|---|---|
| The OS blocks first launch | Follow the OS-specific [launch instructions](docs/安装与启动.md) |
| `npm ci` fails downloading Electron | Check network access and Node version; see [source troubleshooting](docs/安装与启动.md#启动时容易卡住的几件事) |
| The server port is in use | macOS / Linux: `PORT=3801 npm start`; Windows cmd: `set PORT=3801 && npm start`; open the new port |
| Model testing or the task returns an upstream error | Check the API URL, model ID, key and provider quota; see [model setup](docs/配置模型.md) |
| The task waits for an answer or approval | Review the question or approval card above the input, or open the inbox |
| The reply says “done” but the file is missing | Check the active project and results folder; ask it to check `project-update.pptx`. Do not count this attempt as a success |
HTML embedded previews use isolated temporary storage (up to 64 MiB each for `localStorage` and `sessionStorage`, cleared when closed). For a web app that must keep records, choose **Open manually in browser** above the preview. It starts a loopback-only server and opens the current page in your browser, where data survives refreshes under that browser's native storage quota. Keep using the page's own export/backup feature for important data.
Next, try the [core scenarios](#you-say-it-it-hands-you-the-file) or read the [user guide](https://catcatuncle.github.io/openworkbuddy/guides/user.html). Star to bookmark the project; use **Watch → Custom → Releases** for release notifications.
Adjust font size, theme and density from the profile menu → Appearance.
Preview improvements in v0.18.3 fit content automatically on first display and when the preview width changes, while preserving manual zoom. See [preview controls and format limits](docs/文件预览与转换.md) (Chinese; available in v0.18.3).
Intermediate-file cleanup and library deletion now use recoverable trash: the system Trash/Recycle Bin on desktop, or `.trash/files/` with original-path records on a server. Generic output names such as `out/` no longer justify removing an entire directory; active tasks and background commands block cleanup. See [file cleanup and recovery](docs/文件清理与恢复.md) (Chinese). Disk space remains occupied until trash is emptied; native trash failures never fall back to permanent deletion.
When deleting a task, its app-created folder is selected for cleanup by default; uncheck the option to keep its files. Desktop builds move it to the system Trash or Recycle Bin. Server-only builds move it to `.trash/tasks/` under its original workspace, with a `restore.json` recording the original location. Project roots and folders shared by other tasks are retained. Running tasks finish stopping and saving before removal; a cleanup failure keeps the conversation and shows the reason.
<details open>
<summary><b>Your OS blocks the first launch</b></summary>
The current macOS packages use ad-hoc signing and are not Apple-notarized. Use installers from this repository's Releases.
- **macOS, “could not be verified”**: if you choose to proceed, dismiss the notice, then open **System Settings → Privacy & Security → Open Anyway** and complete the system confirmation. See [Apple's instructions](https://support.apple.com/guide/mac-help/open-a-mac-app-from-an-unknown-developer-mh40616/mac).
- **macOS, “is damaged”**: download again first. If it still fails, check the signature and startup log using the [installation guide](docs/安装与启动.md); this is different from an unknown-developer notice.
- **Windows**: after checking the download source, choose “More info → Run anyway” in SmartScreen.
- **Nothing happens**: inspect `~/OpenWorkBuddy/logs/boot.log` and the [startup troubleshooting guide](docs/安装与启动.md#双击了没反应).
</details>
## You say it, it hands you the file
| You say | You get |
|---|---|
| "Build a Q3 review deck from this spreadsheet" | reads → computes → a `.pptx` you can present |
| "Research AI companion apps in China, write a report" | searches → reads each page → Markdown / Word |
| "Turn this material into a page I can read on my phone" | writes HTML → serves it locally → scan the QR |
| "Every day at 9, collect industry news and send it to me on Feishu" | cron + IM push; missed runs catch up |
| "Make a 30-second product promo, vertical and horizontal" | a form for length, aspect and sound → voice-over, motion, compositing → `.mp4` |
<p align="center">
<img src="docs/images/how-it-works.en.svg" width="1190" alt="Ask once; OpenWorkBuddy plans, executes and verifies, and hands back PPTX / DOCX / XLSX / HTML files">
</p>
> [!NOTE]
> Also: parallel tasks, goal-based acceptance, 👍👎 feedback that feeds self-evolution, two-layer memory, permission tiers, remote control over Feishu / WeChat, a desktop pet… Full list (Chinese): **[功能清单](docs/功能清单.md)**.
## Why this one
<table>
<tr>
<td width="50%" valign="top">
<b>📄 The files are real.</b>
Decks, documents, spreadsheets and pages are actually generated — open them from the output panel and check. Review the output panel and the actual file on disk after delivery.
</td>
<td width="50%" valign="top">
<b>🔌 Swap models freely; everything stays yours.</b>
DeepSeek / Qwen / GLM / Kimi / OpenRouter / Ollama switch with one click. Already have <b>Claude Code or Codex</b> on this machine? Use it as the engine — no second token bill. Settings and workspace files remain under your control; cloud models and remote connectors receive the data needed for the task. It listens on <code>127.0.0.1</code> by default.
</td>
</tr>
<tr>
<td width="50%" valign="top">
<b>🧩 Adding a capability = dropping one Markdown file.</b>
Save it as <code>skills/<name>/skill.md</code> and it's live on the next task — no code, no restart, no build.
</td>
<td width="50%" valign="top">
<b>🔍 It's also a readable agent.</b>
Model routing, tool calls, file acceptance, memory, permissions and local traces all live in one repo: for any real task you can see why it did what it did, which model it used, and what it finally handed over.
</td>
</tr>
</table>
<details>
<summary><b>📣 Want to tell someone about it? Copy this</b></summary>
```text
OpenWorkBuddy: an open-source AI office assistant that runs on your own machine.
Ask once; it plans, does the work, checks it, and hands you a PPT / Word / Excel / web page you can open, not a chat log.
· Any model: DeepSeek, Qwen, Kimi, GLM, Ollama, or the Claude Code / Codex already on your machine
· Your data stays local and it listens on 127.0.0.1 by default; or run it on a server for your team, each person binding their own WeChat or Feishu
· Adding a capability = dropping in one Markdown file
Good for: people who live in decks, weekly reports and research docs; teams that want a self-hosted AI assistant; developers who want to see how an agent is built.
Free for personal non-commercial use: https://github.com/CatCatUncle/openworkbuddy
```
</details>
## What it looks like
**"Same person, four different scenes, holding a hand-written sign — make it look like a snapshot, not an AI render."**
<p align="center">
<img src="docs/images/case-photoreal.jpg" width="640" alt="The same person at a cafe window, on a rainy night street, at an office desk and in a morning bedroom, each holding a wooden sign reading 关注 OpenWorkBuddy 项目">
</p>
The hard part isn't drawing a person — it's keeping the same person across all four and the Chinese on the sign legible. So it generates one, then actually looks at what it just made (a real vision call on its own output, not a claim from memory), and fans the rest out from there.
**"Build me a Hunan travel guide site — all 14 prefectures, no skipping."**
<p align="center">
<a href="https://hunan-travel.pages.dev/"><img src="docs/images/case-hunan-site.jpg" width="860" alt="Hunan travel guide: the ink-wash landing screen, and the index of all 14 prefectures grouped by region"></a>
</p>
**<https://hunan-travel.pages.dev/>** — it's live, go click around. One HTML file plus a folder of images, no external CDN. Drop it on any static host and it's a site. Not a mockup — the thing it actually handed over.
**"I'm spending three days in Beijing — plan the trip for me, and use the map card."**
<p align="center">
<img src="docs/images/case-trip-map.jpg" width="860" alt="Map view of a three-day Beijing trip card: day 1's Temple of Heaven, Tiananmen, the Forbidden City, Jingshan and Qianmen pinned in order, with photo place cards on the right and “Open route” and “Send to phone” at the top">
</p>
<p align="center">
<img src="docs/images/case-trip-route.jpg" width="860" alt="Day 1 after “Open route”, switched to Google Maps: the Temple of Heaven, Tiananmen Square, the Palace Museum, Jingshan Park and Qianmen Street joined into one route in order, under the same names as on the trip card">
</p>
<p align="center">
<img src="docs/images/case-trip-timeline.jpg" width="860" alt="The same trip card switched to the timeline: day 3 from the Summer Palace in the morning to Nanluoguxiang">
</p>
The itinerary in the answer becomes a trip card: switch between days, every stop is pinned on the map in order, and the place cards on the right come with photos. "Open route" carries the day's stops into the map in order, under the same names as on the card; in China it opens AMap by default with a switch to Google (that's the second shot), abroad it opens Google Maps; "Send to phone" shows a QR code to carry on from your phone. Flip to "Timeline" at the top right and the day runs from morning to evening. Works without a Key; add an AMap Key in Settings → Map and places in China get more accurate.
**"Every morning at seven, send me today's weather and what I should watch out for, on Feishu."**
<p align="center">
<img src="docs/images/case-schedule-feishu.jpg" width="820" alt="A scheduled OpenWorkBuddy task pushing a morning briefing to Feishu: today's weather and UV, tomorrow's rain, plus two things worth doing today">
</p>
Scheduled tasks run while the application is running. Automation → Run history shows execution and notification delivery separately. Feishu notifications require bot credentials and an explicit recipient ID in Assistant settings; see [IM and scheduled tasks](docs/IM与定时任务.md).
How it pulled those off → **[三个案例，拆开讲](docs/案例.md)** (Chinese, but the screenshots speak for themselves)
## Creative infinite canvas
Start with an image, video or set of assets. Describe, connect references, generate and continue with variations or video directly from each card. Short-drama workflows also organize scripts, characters, scenes, shots, voice and the edit timeline. Connections become real generation inputs.
Edit character and location descriptions and choose an image model directly on their cards. Unspecified models follow the current app’s default; Agent writes with unregistered model IDs are rejected. Invalid model IDs are caught before submission; choosing the default overrides an inherited script model. [Usage and troubleshooting](docs/无限画布使用.md#角色定妆照与场景图).
Ark portrait errors stay on the card. Use approved asset IDs for reference generation; video frame uploads preserve original bytes. [Portrait assets and generation modes](docs/无限画布使用.md#火山视频提示输入图片可能包含真人).
<p align="center">
<img src="docs/images/canvas-start.png" width="960" alt="OpenWorkBuddy canvas example: start an image or video, import assets and open Agent when needed">
</p>
<details>
<summary>Example: a 75-second AI short-drama episode</summary>
**A whole AI short-drama episode from one sentence**: ask for "an episode of 《不烧心》 as a vertical short drama" and it goes script → character sheets → storyboard → first frames → shot videos → final cut. This 75-second episode was made that way (click the cover to watch).
<p align="center">
<a href="docs/videos/ai-short-drama-ep01.mp4"><img src="docs/images/ai-short-drama-poster.jpg" width="300" alt="Cover of the AI short drama 《不烧心》 EP01, vertical cut — click to play the 75-second episode"></a>
</p>
</details>
Open "Infinite canvas" in the left sidebar. Drag empty space to pan, `Shift`+drag to marquee-select, `Shift`/`⌘`+click to add or drop nodes from the selection, and expand Agent to reference nodes or assets with `@`.
Edit models, references and prompts on cards, compare versions and open Agent when needed. Empty prompts never start generation. Growing cards move their neighbors once; compact result cards retain visible failure details. Duplicate with references, arrange selections, navigate with the optional minimap and export the canvas together with its assets. See the [canvas guide](docs/无限画布使用.md) (Chinese).
## Put it on a server for your team: multi-user, self-hosted
```bash
git clone https://github.com/CatCatUncle/openworkbuddy.git && cd openworkbuddy
bash deploy.sh --domain buddy.example.com # Docker + automatic HTTPS
```
Multi-tenant, with seats, quotas and one-click offboarding. **Register the admin first** — the first account becomes super admin, so keep that password safe. Members cannot read company keys, and the admin console decides whether they may add their own models. Direct Linux installs need bubblewrap first.
→ [部署](docs/部署.md) · [ops handbook](deploy/README.md) · [多人协作](docs/多人协作.md) · [账号与部署运维](docs/账号与部署运维.md) (Chinese)
## Models: DeepSeek, Qwen, Kimi, Ollama, Claude Code, Codex
**Settings → Models**: pick a preset (OpenAI / Anthropic / OpenRouter / DeepSeek / Qwen / Zhipu / Kimi / Volcengine Ark / Ollama), paste the key, save — it applies immediately. Image, voice, transcription and video models have their own table, tried in order when several are set; retired models warn ahead of shutdown. Reference → [配置模型](docs/配置模型.md) (Chinese)
Assistant mode can also select local Codex, Claude Code, or a configured API model. It follows your current engine by default; changing its selection does not change scheduled task settings.
> [!IMPORTANT]
> Keys live only in `config.json`, which is already in `.gitignore`. Don't commit it.
## Command line
`openworkbuddy` shares **one** set of config, skills, memory, connectors and sessions with the desktop app — start something in the terminal and you can watch it and chime in from your phone; stop halfway on the desktop and `openworkbuddy resume` picks it up.
Inside a source folder it uses a lean coding profile and edits in place. The default display keeps one progress line, the final answer and files actually changed in the current turn. Use `/details` for tool output, `/usage` for token usage, and `/verbose on` for detailed activity. Say “open it” after creating a page to open the session's result with a local application, or use `/open 1` and `/open subdirectory/index.html`.
```bash
npm link # once: install the global command (or just run `node cli.js …`)
openworkbuddy "write my weekly report" # one-shot: runs and exits
openworkbuddy # interactive: type / for the command menu
cat error.log | openworkbuddy "what is this" # pipe: stdin becomes material
openworkbuddy -q "write my weekly report" > report.md # just the report, no progress bars
```
One-shot and pipe mode **never** ask you questions, so scripts and cron don't hang. Exit codes mean something: `0` success, `1` task failed, `2` bad arguments, `130` Ctrl+C — so `openworkbuddy doctor && npm start` stops a misconfigured machine before it starts.
`sessions` / `resume` / `engines` / `doctor` / `pair` (QR-pair your phone) / `worktree`, plus `--mode` `--perm` `-C` `-f` `--json` and the rest → **[命令行用法](docs/命令行用法.md)** (Chinese)
## How it's put together
<p align="center">
<img src="docs/images/architecture.en.svg" width="1120" alt="OpenWorkBuddy architecture: desktop, CLI and IM clients enter through their hosts into the Agent Runtime, which coordinates models, safety, tools, project context, deliverables and traces">
</p>
Entry points are `server.js`, `cli.js` and `src/im/`; the Agent is in `src/agent/agent.js`, with tools in `src/agent/tools.js` and `src/tools/`. Sessions, config and usage are stored locally under `data/`; task files go to the project workspace. See [code architecture](docs/代码架构.md) for layers and [implementation details](docs/实现细节.md) for a task walkthrough (both Chinese).
The built-in Agent alternates model decisions and tool execution, with concurrent read-only batches and independent media generation. Progress improvements in v0.18.3 periodically reconcile checklists, separate confirmed completion from current activity, and keep expert checklists from replacing the main plan. Tool counts never become completion percentages. See [Agent execution and progress](docs/Agent执行与进度.md) for the comparison with Plan / Execute and LLMCompiler (Chinese; available in v0.18.3).
## What's new
- **Nov 10** v0.19.3: attachment names with a Chinese closing parenthesis are no longer cut short, so conversations stop showing a broken extra attachment.
- **Nov 6** v0.19.2: library tabs no longer stay highlighted when Notes is open, so only one tab is selected at a time.
- **Nov 6** v0.19.1: multi-user Linux runs member commands in a system sandbox, closing a blocklist bypass that could read key files. Install bubblewrap before upgrading. See [deployment](docs/部署.md).
- **Nov 6** The admin console controls which connectors and skills members can use, sets storage quotas and turns on in-organization task sharing. Connector setup gains built-in checks.
- **Nov 6** Memories can belong to a project, so one client's terms stay out of another client's tasks.
- **Nov 6** Press / on the canvas to search nodes. Split images into grids, make videos in batches, and save card groups as reusable workflows. See [canvas](docs/无限画布使用.md).
- **Nov 6** The engineering lane shows the working folder and change count, with per-file diffs and discarding one file's edits. Folders with source code run in place.
- **Nov 6** The library responds immediately with tens of thousands of files because scans moved to a background thread. Concurrent tasks queue by available memory, and heavy rendering has its own limit.
- **Nov 6** CLI coding sends 75% less prompt text per turn, runs in place inside source folders, and loads extra tools on demand.
- **Nov 6** Eval sets can be uploaded as CSV, Excel or TXT, or built in the app, with weights, required items and penalties. Human and judge agreement reports Cohen's kappa.
- **Nov 6** Retired Ark models warn ahead of shutdown; image and video generation fall back in order. Doubao voice, transcription and custom speech APIs are added; test them against your account first.
- **Nov 6** Hover a message to see its time, jump from a reply to its trace and back, and read TTFT and TPOT in traces. The sidebar lists projects directly without pushing tabs away.
- **Nov 4** v0.18.4: model setup: choose cloud model IDs, refresh key-specific lists, and diagnose retired or unauthorized Ark endpoints. See [configuration](docs/配置模型.md).
- **Nov 3** v0.18.3: recoverable file cleanup protects output directories and active tasks; previews fit automatically on open, and long tasks reconcile progress during execution. See [cleanup and recovery](docs/文件清理与恢复.md) (Chinese).
Format support and limits: [file previews and conversion](docs/文件预览与转换.md). Full history: [CHANGELOG](CHANGELOG.en.md).
## ⚠️ This agent has a shell
It runs commands, edits files and goes online, so there are command approvals, file blocklists, a URL allowlist, an audit log and four permission levels; on macOS and Windows its commands also run in a system sandbox that can't read your Keys or ledger; multi-user Linux uses bubblewrap and refuses member commands if isolation cannot start. Third-party skills get a static check before install and risky ones are refused by default — but it isn't antivirus; read the `skill.md` yourself.
**Read [安全](docs/安全.md) before exposing it to the internet** — defaults are tuned for local use.
## Contributing
- **Something broke? [Open an issue](https://github.com/CatCatUncle/openworkbuddy/issues/new)**, even if it's one line of error text. Scrub your API keys first.
- **10 minutes** — write a skill: one Markdown file at `skills/<name>/skill.md`, live on save. [Template](CONTRIBUTING.md#提交一个技能3-分钟)
- **One evening** — pick an issue: `npm install && npm start` runs it, `npm test` goes green without any API key
Project layout, tests and PR conventions are in [CONTRIBUTING.md](CONTRIBUTING.md). No need to open an issue first — send the PR.
## Coming from another tool: open-source alternative to Claude Cowork, Codex and more
Looking for an **open-source alternative to Claude Cowork, Codex, Doubao or Qwen office features**, or a **DeepSeek agent harness**? Each has an honest comparison page, including when the other product is the better fit:
| You use / you're looking for | Comparison |
|---|---|
| Claude Cowork | [Open-source Claude Cowork alternative — self-hosted, any model](https://catcatuncle.github.io/openworkbuddy/alternatives/claude-cowork/) |
| Codex | [Codex alternative for office work — and it can drive the Codex CLI](https://catcatuncle.github.io/openworkbuddy/alternatives/codex/) |
| Doubao | [Doubao office alternative — still runs Doubao models](https://catcatuncle.github.io/openworkbuddy/alternatives/doubao/) |
| Qwen | [Qwen office alternative — runs Qwen via DashScope or Ollama](https://catcatuncle.github.io/openworkbuddy/alternatives/qwen/) |
| DeepSeek | [DeepSeek agent harness — paste a key and it works](https://catcatuncle.github.io/openworkbuddy/alternatives/deepseek-agent/) |
OpenWorkBuddy is an independent open-source project, not affiliated with any of these companies.
## Docs
First-success testing, timing definitions and missing metrics → [Adoption and feedback](docs/采用与反馈.md#english).
Most docs are in Chinese; the code and comments are the source of truth.
Online guides: **[User guide](https://catcatuncle.github.io/openworkbuddy/guides/user.html)** · **[Developer guide](https://catcatuncle.github.io/openworkbuddy/guides/developer.html)** · [Bilingual project website](https://catcatuncle.github.io/openworkbuddy/)
| Doc | What's in it | Doc | What's in it |
|---|---|---|---|
| [功能清单](docs/功能清单.md) | Every capability, skill and tool | [部署](docs/部署.md) | Server / Docker / reverse proxy |
| [案例](docs/案例.md) | How the pictures above were made | [多人协作](docs/多人协作.md) | Multi-tenant, accounts, quotas |
| [安装与启动](docs/安装与启动.md) | Installers, source, common snags | [安全](docs/安全.md) | Approval gates, allowlists, audit |
| [配置模型](docs/配置模型.md) | base_url / model names per provider | [数据同步与搬家](docs/数据同步与搬家.md) | Where data lives, moving machines |
| [命令行用法](docs/命令行用法.md) | CLI flags, pipes, `--json`, cron | [开源与商业版边界](docs/开源与商业版边界.md) | What a licence actually buys |
| [扩展](docs/扩展.md) | Skills, MCP, plugins, experts | [路线图](docs/路线图.md) | What's next, what counts as done |
| [IM与定时任务](docs/IM与定时任务.md) | Feishu / QQ / WeCom / WeChat / DingTalk | [实现细节](docs/实现细节.md) | How the agent loop actually runs |
| [安全基线](docs/安全基线.md) | Where data lands, who can read it, what isn't covered | [远程访问](docs/远程访问.md) | Reaching your machine from outside; both switches off by default |
| [账号与部署运维](docs/账号与部署运维.md) | Accounts, offboarding, password recovery, backups, running several orgs | [代码架构](docs/代码架构.md) | How the code is layered, who may require whom, where new code goes |
## Also by the same author
- **[toolward](https://github.com/CatCatUncle/toolward)** — a static safety check for agent skills and MCP connectors: 37 rules in six families, zero runtime dependencies, Node 20.10+. Run `npm i -g toolward` and OpenWorkBuddy picks it up as a second ruler automatically; skip it and nothing changes. Like this project it is PolyForm Noncommercial.
## Commercial license · Enterprise services
> [!IMPORTANT]
> **Free for personal non-commercial use, schools and non-profits. Company use (even internal only) and personal commercial use require a paid commercial license.**
> Companies may evaluate it free for up to 30 days, as long as it isn't used in real business yet; long-term use requires a license.
> **To purchase a commercial license, email [contact@aijentra.com](mailto:contact@aijentra.com) or add WeChat `agi_smith`.** Include your company, use case and expected user count; mention “OpenWorkBuddy commercial license” in your WeChat request. Terms: [COMMERCIAL-LICENSE.md](COMMERCIAL-LICENSE.md).
| Your situation | License needed? |
|---|---|
| Personal non-commercial use (learning, research, hobby projects) | Free |
| Schools, research institutions, non-profits, government bodies | Free |
| Company evaluation: up to 30 days, not yet used in real business | Free |
| A company or team using it in real business, **even internal only** | **Required** |
| **Personal commercial use**: freelancing, client work, paid products or services | **Required** |
| SaaS offered to others, bundling into products you sell, client delivery / outsourcing / custom work | **Required** |
A license unlocks no features: there is only this one codebase, with no feature locks, no countdowns and no online license checks. The software sends no usage data to the author or any third party; when it checks for updates, it asks GitHub Releases whether a new version is out (GitHub can see your IP).
Commercial use without a license can be pursued by the rights holder under civil law, and removing or altering copyright notices without permission may itself infringe — legal basis and court cases in [未授权商用的法律后果](COMMERCIAL-LICENSE.md#未授权商用的法律后果) (Chinese).
Beyond licensing, we take on enterprise AI work:
| Service | What it covers |
|---|---|
| **Enterprise FDE (forward-deployed engineering)** | Engineers join your team, wire Agents into real business workflows, and hand over once it runs |
| **AI training** | In-house training so your team uses AI in everyday work, not just demos |
| **AI adoption consulting** | Pick the use cases, shape the plan, size the cost; private deployment and industry agents included |
The author is a former big-tech Agent engineer who has delivered AI solutions for cross-border e-commerce, manufacturing, AI startups, private funds, major consumer brands and state-owned enterprises. Based in Shenzhen — visits welcome.
📮 Licensing and business inquiries: **[contact@aijentra.com](mailto:contact@aijentra.com)** · WeChat: **`agi_smith`**
## License
**Free for personal non-commercial use, schools and non-profits; company use (even internal only) and personal commercial use need a commercial license.** Companies may evaluate free for 30 days. To purchase a commercial license, email **[contact@aijentra.com](mailto:contact@aijentra.com)** or add WeChat **`agi_smith`** (mention “OpenWorkBuddy commercial license”).
[PolyForm Noncommercial 1.0.0](LICENSE); terms and what unlicensed commercial use leads to are in [COMMERCIAL-LICENSE.md](COMMERCIAL-LICENSE.md). Deploy configs, scripts and skill templates are also MIT ([the list](COMMERCIAL-LICENSE.md#按-mit-授权的部分)); the name and logo aren't licensed.
Anyone who copies or redistributes any part of this software must keep the license terms (or their URL) and every line beginning with `Required Notice:` at the top of [LICENSE](LICENSE).
Official releases are published only at CatCatUncle/openworkbuddy on GitHub and its Releases. Installers from other sources get no updates or support from the author; users' rights are those under the PolyForm Noncommercial License 1.0.0 and any commercial license.
**KylinWork and others built on this project's code without the author's permission and changed its license; the author has demanded they be taken down everywhere.**
Copyright (c) 2026 开发者猫叔 (DeveloperCatUncle)
## Disclaimer
An independent open-source project, not affiliated with Tencent or its WorkBuddy product, containing none of its code or assets; IM integrations use public APIs only. Borrowed ideas are credited in [NOTICE.md](NOTICE.md). Concerns → [Issues](https://github.com/CatCatUncle/openworkbuddy/issues).
## Support this project
<p align="center">
<a href="https://github.com/CatCatUncle/openworkbuddy">
<img src="docs/images/star-guide.svg" width="640" alt="That Star button in the top-right corner — press it">
</a>
</p>
<p align="center">
<a href="https://github.com/CatCatUncle/openworkbuddy"><img src="https://img.shields.io/github/stars/CatCatUncle/openworkbuddy?style=for-the-badge&logo=github&label=Star%20this%20repo&color=5b5ff7" alt="Star this repo"></a>
</p>
<p align="center">
<sub>Pass it to one colleague who hand-builds decks, weekly reports and meeting notes — worth more than a hundred impressions.</sub>
</p>
## Contributors
Thanks to everyone who has changed something here. Want to join them: [CONTRIBUTING.md](CONTRIBUTING.md).
<p align="center">
<a href="https://github.com/CatCatUncle/openworkbuddy/graphs/contributors">
<img src="https://contrib.rocks/image?repo=CatCatUncle/openworkbuddy" alt="OpenWorkBuddy contributors">
</a>
</p>
## Star history
<p align="center">
<a href="https://star-history.com/#CatCatUncle/openworkbuddy&Date">
<picture>
<source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/svg?repos=CatCatUncle/openworkbuddy&type=Date&theme=dark">
<img src="https://api.star-history.com/svg?repos=CatCatUncle/openworkbuddy&type=Date" alt="Star History Chart" width="600">
</picture>
</a>
</p>