Rich, inventive capabilities and serious productivity gains —
with all your data staying on your own machine. Safe for individuals and enterprises, free to commercialize, open to audit.
English · 简体中文
🌐 Website · 📖 Docs · 🚀 Get started · 📦 Releases · China mirror · 发版手册 · 💬 Community
Important
Public-interest open-source and data-risk notice: OpenHub is a public-interest open-source project. The maintainers do not assume responsibility for user data loss, corruption, or unrecoverable damage during iteration. Back up your data before upgrades, migrations, experimental features, or real production use.
OpenHub is everything you imagine an AI workstation to be — and it runs on your terms. One React frontend and one Rust backend give you an evolving desktop companion, an unattended automation platform, a unified knowledge base, native computer- and browser-use, and an open capability bus that any agent can drive. No cloud account. No telemetry. No subscription. Your data never leaves your machine except for the LLM calls you configure.
The product name is OpenHub. Lowercase openhub is used only for code identifiers, crate names, environment variables, and repository paths.
| 🔓 Open & local | Source fully open, no reservations. Data lives on your machine and is never sent out on its own. Free for personal and commercial use. Open to audit. |
| 🐾 Evolving companions | The most complete companion-growth system we know of — it learns how you work and gets better over time. Not just a buddy, a genuine productivity partner. |
| 🤖 Unattended automation | Manage requirements, then just give the order. AutoWork + IDMM keep your sessions alive and working reliably while you're away. |
| 🌐 Open capability ecosystem | Everything is here, everything connects, everything cooperates — and any agent can borrow OpenHub's powers over MCP / REST. |
| 🧩 Config once, use anywhere | Unified management of knowledge bases, skills, agents, MCP servers, and models — defined once, reused across every surface. |
| 🖥️ Truly native | In-process, self-built computer use and browser use as native tools — more capable, faster, and cheaper on tokens. |
| 🚀 Built for productivity | Designed from real needs, with a lot of inventive capabilities. And many delightful features are still on the way. |
Data security is not a setting in OpenHub — it is the architecture.
- All data is local. OpenHub never proactively sends your data anywhere. The only outbound network calls are the LLM requests you explicitly configure to your chosen model provider. There is no other third-party service integration phoning home.
- Safe for anyone who cares about data. Individuals and enterprises with strict data-handling requirements can use it with confidence. The code is fully open and open to audit.
- We cut features to keep this promise. To guarantee your data stays yours, we deliberately dropped several advanced, genuinely fun feature designs. Everything here is in service of letting users — and developers — relax.
- No ads. No commercialization. No membership tiers. We promise to never charge for any feature of this project. The only thing that costs money is your LLM provider's tokens, which is outside our control. (If finding/serving models is painful, reach out — we're happy to help build a unified model gateway.)
See SECURITY.md for the deployment threat model and responsible-disclosure policy.
🎬 Demo videos: China: Bilibili | International: YouTube
Work orchestration: conversation, reusable roles, and DAG canvas
 Desktop companions · memory and growth |
 Requirements platform · AutoWork entry |
 Knowledge base · local domain context |
 Terminal · capabilities for Agent CLI |
 Assistants & Skills · scenario capability library |
 Models & Agents · unified management and setup |
Captured from the live OpenHub desktop app on 2026-07-01 and kept at 2560px wide. See the screenshot manifest for the full set and capture method.
Guide:
docs/guides/companions.md
The companion you talk to every day quietly becomes the assistant who gets you.
- Make it yours. Upload a custom companion figure (DIY), or pick from an independent figure library decoupled from any single companion.
- One brain, many faces. Run multiple companions that share a common memory hub, while each keeps its own private memory and can mount different domain knowledge bases. Teach one companion well, then have it teach the others.
- Chat with them where you already work. Companion chats now live in the main Sessions UI under a dedicated desktop-companion group, while
/openhubstays focused on companion management. - It learns you (opt-in, on by default after a one-time consent). A background learner distills your usage into durable memories; a deterministic evolution engine mines your recurring multi-step tool sequences into draft skills it proposes for your review. Memory is fully visible and editable.
- Skills that spread. Companions generate their own skills, discuss them with you, and can gift a skill to another companion (the recipient gets a copy) — opt-in shared learning across your roster.
- A super gateway, not just a buddy. Each companion is a complete, independent individual that can connect to multiple IM channels. From anywhere with a network and a chat app, message your companion to drive your computer for you. Each companion can fully operate the desktop's capabilities.
Start from a normal conversation, then let OpenHub expand the job into a live DAG when the task deserves it.
- Conversation first. Launch multi-agent work from the same chat you are already using, then open the orchestration rail / floating canvas without leaving that conversation.
- Per-node preflight control. Before a worker starts, override its model and add a preset brief; settled nodes can be rerun with the same pre-configuration.
- Review before execution. Interactive runs stop at plan-ready and surface an in-conversation approval banner, so you can adjust the DAG before work begins.
- Project real worker transcripts. Click any node to read that worker's actual conversation in the main content area, then jump back to
mainto keep steering the lead agent.
You give the orders; OpenHub reliably does the work.
- Requirement platform — a CRUD store with ordered rotation, a board/kanban, tags, and per-item claim.
- AutoWork — claims pending requirements, drives a turn, rotates to the next, and renews leases while a turn is in flight. Targets can be conversation agents or terminal PTYs.
- IDMM (Intelligent Decision-Making) — per-session supervision that keeps agents alive through provider faults and decision stalls, with a no-LLM rule tier and a sidecar backup-model tier, stacking on top of AutoWork.
- Notify out — completion notifications to Lark/Feishu custom bots, Slack, and HTTP webhooks.
Pull the knowledge scattered across your system into one managed, trackable place.
- Centralized management & tracking — create, mount, and track consumers across conversations, terminals, and companions.
- Safe write-back — a code-enforced, per-surface write policy. By default, writes are staged into a review inbox with unified-diff preview and merge/discard — so agents never scribble into the wrong place.
- Real-time URL snapshot — turn any web page into a knowledge source (SSRF-guarded fetch, HTML→Markdown), in snapshot (persisted, re-fetchable) or live mode.
- Scoped retrieval — agents call a
knowledge_searchtool whose scope is decided server-side and cannot be widened.
Self-built, in-process Rust — no Playwright, no Node, no third-party automation daemon. More capable, faster, and far cheaper on tokens, with fine-grained control and fully open source for you to extend.
- Computer use — accessibility tree + Set-of-Marks overlay + OCR, steering the model to act on real UI elements instead of guessing pixels. macOS (AXUIElement + Vision OCR) and Windows (UI Automation) are complete; Linux (AT-SPI2) is partial.
- Browser use — an in-process Chromium CDP engine with ARIA observation, an egress firewall with out-of-band approval, and an origin-bound secret vault so credentials never reach the LLM.
- Modes that match the job — desktop builds default to silent browser runs in the bundled Chromium, but you can switch to your own system Chrome / Edge when site compatibility matters.
- One-click login reuse — open a visible "Log into my browser" window once, then later silent agent runs reuse that authenticated state from a shared profile with encrypted backup.
- Approval with visual context — silent-mode high-risk browser approvals can include a current-page screenshot, so you do not have to approve blindly from text alone.
- Guarded by design — every action carries a danger × surface approval matrix; irreversible actions wait for explicit confirmation.
ℹ️ Computer/browser control ship with the desktop app. The headless web/server host omits them by design.
Guides:
remote-capability-api.md·remote-capability-api-examples.md
Every capability OpenHub has is exposed through a single, typed capability registry — ~20 domains and 150+ tools — so you can wire OpenHub into anything.
- MCP front door at
/mcp(authenticated, Streamable-HTTP). Point Claude Code, Cursor, or your own agent at it and they operate OpenHub exactly as the desktop companion does. - REST + OpenAPI at
/v1/tools, with streaming and an auto-generated/v1/openapi.json. - Adding a capability to the bus makes it appear on MCP and REST automatically — no drift.
Guide:
docs/guides/model-routing.md
- Built-in
openhubagent — no extra install. Works with 26+ model providers/presets (OpenAI, Anthropic, Gemini + Vertex AI, AWS Bedrock, DeepSeek, OpenRouter, Moonshot/Kimi, Qwen/Dashscope, Zhipu/GLM, MiniMax, SiliconFlow, xAI, Volcengine/Doubao, and more) across 4 wire protocols, plus the New API aggregator gateway. - ~19 external agents over ACP — connect Claude Code, Codex, Gemini, Qwen, Kimi, Cursor, Copilot, Goose, OpenCode, Droid, and more, and OpenHub feeds them models and its native capabilities (computer/browser/knowledge/gateway) over injected MCP bridges.
- Everywhere — the native capabilities are available to the built-in agent, to ACP agents, in the chat UI, and in the terminal.
- Graceful multimodal fallback — if a selected provider/model rejects image input, OpenHub strips the images, retries in the same conversation, and leaves an inline notice instead of killing the session.
- Per-model context tuning — override context-window limits per model when an upstream platform reports bad defaults or hides them, improving routing and long-context budgeting.
OpenHub does not lock you into a single model vendor. Pick providers by region, price, quota, model capability, and data policy, then paste the API key into OpenHub's Models & Agents page. These are third-party services; pricing, regional availability, rate limits, and data-handling terms are controlled by each provider.
| Provider | Start here | Good to evaluate |
|---|---|---|
| Platform | Step models for Chinese, agentic, and cost-conscious workloads | |
| API keys | Long context, Chinese writing, coding, and general tasks | |
| API keys | GLM models, general reasoning, coding, and enterprise integration | |
| API keys | Doubao models and China-region cloud/enterprise workflows | |
| API keys | Qwen models, DashScope, and Alibaba Cloud workflows | |
| API keys | MiniMax models, long-form text, multimodal, and voice capabilities | |
| Website | MiMo models and Xiaomi ecosystem capabilities | |
| API keys | Reasoning, coding, and high value-for-money model calls | |
| API keys | Multi-model aggregation, unified billing, fallback routing, and comparison | |
| API keys | Claude models, long-form work, coding, and the Claude Code ecosystem | |
| GPT models · API keys | GPT models, OpenAI API, agent workflows, coding, and general-purpose tasks | |
| API keys | Gemini models, multimodal work, very long context, and Google AI Studio |
Guide:
docs/guides/terminal.md
Run agent CLIs inside in-app PTY sessions (or the standalone openhub CLI). OpenHub injects native capabilities — knowledge search, requirement completion, and lifecycle hooks — into known CLIs through their own native config, so you keep full fidelity and OAuth.
No social platform required. One-tap QR pairing connects your phone or tablet to your computer over the LAN (one-time token, realtime over WebSocket) so you can drive your workstation remotely from the couch.
Central hubs for Knowledge, Assistants & Skills, MCP, Models, and Open Capabilities — define them once, then select per conversation, terminal, channel, or companion. One source of truth, reused everywhere.
Guide:
docs/guides/channels.md
Bind a companion to any of these and drive it from where you already chat:
Telegram · Lark / 飞书 · DingTalk / 钉钉 · WeChat / 微信 · Discord · Slack · Matrix · Mattermost · Twitch · Nostr · QQ Bot
One React frontend, one Rust backend, two host modes — and the same backend runs in-process in both.
openhub-desktop |
openhub-web |
|
|---|---|---|
| Shell | Tauri 2 desktop app | Standalone axum server |
| Backend | Embedded in-process, private loopback port | Same backend, in-process |
| Auth | Local-trust token injected into the webview | Login required by default |
| Serves | Native desktop UI + tray + companion windows | API + /ws + built SPA on one port |
| Computer / browser use | ✅ Included | ❌ Headless (omitted) |
There is no Electron shell, no Node web host, and no prebuilt backend handoff.
Repository map
apps/
desktop/ Tauri 2 shell and desktop-only commands
web/ standalone web host for API + SPA
crates/
agent/ 15 openhub-* crates: engine, providers, tools, MCP, skills, memory,
browser/computer use, and the standalone openhub CLI
backend/ 29 openhub-* crates: app composition, auth, database, sessions,
MCP, knowledge, requirements, terminal, companion, gateway, etc.
shared/ 2 cross-layer crates: openhub-net and openhub-redact
ui/ React 19 + Vite SPA shared by desktop and web
docs/ technical docs, user/operator guides, architecture notes
packaging/ Linux deployment support for the web host
Start with docs/architecture/overview.md for the full system map. The Cargo workspace is defined in Cargo.toml.
📦 Installers: use GitHub Releases first. Mainland China users can use the Baidu Netdisk mirror (shared as openhub). You can also install from source or run the server with Docker.
Prerequisites
- Rust — stable toolchain, edition 2024
- Bun ≥ 1.3.13
- Recommended on PATH for full agent tooling:
node/npm/npx,git,ripgrep
Desktop app (from source)
git clone https://github.com/Grong/openhub-client.git
cd openhub-client
bun install
bun run dev # develop with hot reload
bun run build # build a desktop bundle for your OSWeb server (self-host)
bun run build:ui && bun run serve:web
# serves API + SPA on http://127.0.0.1:8787 (login required)Docker (self-host the server)
docker compose up -d --build
# then open http://<server-ip>:8787 — pair with the bundled Caddyfile for TLSSee docs/getting-started/installation.md and docs/guides/web-server-deployment.md for details.
bun install # install dependencies (one-time)
bun run dev # desktop app development (hot reload)
bun run dev:web # web host + Vite development
bun run build:ui # build the SPA
bun run check # frontend typecheck + i18n + theme + script-registry gate
bun run test # Rust tests (use test:fast for nextest)Prefer the scripted entry points over plain cargo/vite — they include build-dir pruning and consistency checks. New to the codebase? Read CONTRIBUTING.md, CONTRIBUTING.zh-CN.md, and docs/contributing/development.md.
Each OS has its own command, and a package can only be built on its matching OS —
macOS bundles must be signed/notarized on macOS, Windows installers built on Windows,
Linux packages on Linux. There is no cross-OS build. All artifacts are collected into
dist/desktop/.
Common argument shape for all three:
bun run build:<os> [arch ...] [--signed] [-- <args passed straight to `tauri build`>]
- arch — zero or more architectures. Omit to use the per-OS default below.
--signed— sign (and, on macOS, notarize). Requires local signing config; see each OS.-- …— everything after--is forwarded verbatim totauri build(e.g.-- --bundles nsis).build:macandbuild:winalso forward unknown--xxxoptions directly. For updater builds, layer on the committed overlay as a file path:bun run build:<os> --config apps/desktop/tauri.updater.conf.json— pass the file, not inline JSON, because Windows PowerShell 5.1 strips the quotes from--config '{...}'.
macOS — build:mac (produces .dmg; default arch: universal)
| Goal | Command |
|---|---|
| Universal (Intel + Apple Silicon, one fat package) | bun run build:mac |
| Universal, signed + notarized | bun run build:mac --signed |
| Apple Silicon only | bun run build:mac arm |
| Intel only | bun run build:mac intel |
| Intel only, signed + notarized | bun run build:mac --signed intel |
| All three separately (ARM + Intel + Universal) | bun run build:mac arm intel universal |
Arch aliases: arm/aarch64/silicon, intel/x64/x86_64, universal/all-arch.
Signing reads apps/desktop/signing/.env.signing (gitignored); missing → it errors with setup hints.
Windows — build:win (produces a single NSIS .exe; default arch: the host's, usually x64)
| Goal | Command |
|---|---|
| Current arch | bun run build:win |
| x64 only | bun run build:win x64 |
| ARM64 only | bun run build:win arm64 |
| Both | bun run build:win x64 arm64 |
| Signed (Authenticode) | bun run build:win --signed |
Arch aliases: x64/x86_64, arm64/aarch64/arm. --signed reads the cert thumbprint from
WINDOWS_CERTIFICATE_THUMBPRINT (no “notarization” concept on Windows).
Linux — build:linux (produces .deb / .AppImage / .rpm; default arch: the host's)
| Goal | Command |
|---|---|
| Current arch | bun run build:linux |
| x64 only | bun run build:linux x64 |
| ARM64 only | bun run build:linux arm64 |
| Both | bun run build:linux x64 arm64 |
Arch aliases: x64/x86_64, arm64/aarch64/arm. Linux has no signing/notarization step.
bun run buildstays as the simple "just build for whatever OS I'm on" shortcut; thebuild:<os>commands above add explicit arch selection, signing, anddist/desktop/collection.
Full script catalog
| 脚本 | 说明 |
|---|---|
| 开发（热重载） | |
bun run dev |
启动桌面应用开发（tauri dev，热重载） |
bun run dev:web |
启动 Web 全栈开发（后端 API + 前端 vite） |
bun run dev:ui |
仅启动前端开发服务器（纯 vite，无后端） |
| 构建（出制品） | |
bun run build |
为当前操作系统打桌面安装包 |
bun run build:win |
打 Windows 安装包（NSIS），汇总到 dist/desktop/ |
bun run build:mac |
打 macOS 安装包（.dmg），汇总到 dist/desktop/ |
bun run build:linux |
打 Linux 安装包（.deb/.AppImage/.rpm），汇总到 dist/desktop/ |
bun run build:signed |
打桌面包并签名+公证（仅 macOS） |
bun run build:updater |
打桌面包并产出自更新 .sig 制品 |
bun run make:latest |
扫描本机更新产物，生成/合并自动更新清单 latest.json |
bun run release:mac |
一键 macOS 发版：自动判定追加/首发；首发用 -Version 打版本号 + -NotesFile/-Notes 建 Release；-DryRun 只预检 |
bun run release:win |
一键 Windows 发版：自动判定追加/首发；首发用 -Version 打版本号 + -NotesFile/-Notes 建 Release；-DryRun 只预检 |
bun run release:linux |
一键 Linux 发版：自动判定追加/首发；首发用 -Version 打版本号 + -NotesFile/-Notes 建 Release；-DryRun 只预检 |
bun run build:ui |
前端生产构建 → ui/dist |
| 运行（组装好的应用） | |
bun run serve:web |
启动 Web 服务器，托管已构建的前端 |
| 测试 | |
bun run test |
运行全部 Rust 测试（含 doctest） |
bun run test:fast |
用 nextest 快速跑 Rust 测试（日常） |
| 静态检查 / 门禁 | |
bun run check |
聚合静态门禁：typecheck + i18n + 主题契约 + 图标导入 + 脚本登记 |
bun run typecheck |
前端 TypeScript 类型检查（tsc --noEmit） |
bun run check:i18n |
校验 i18n 类型与 locale 键是否一致 |
bun run check:theme |
校验预设 CSS 主题契约 |
bun run check:icons |
校验 @icon-park/react 导入禁别名/禁命名空间（别名会被图标包装插件改写成非法代码，tsc 抓不到） |
| 格式化 | |
bun run fmt |
格式化 Rust 代码（cargo fmt） |
bun run fmt:check |
校验 Rust 代码格式（cargo fmt --check） |
| 代码生成 | |
bun run gen:i18n |
由 locale 重新生成 i18n 类型声明 |
| 维护 / 工具 | |
bun run clean |
深度回收构建空间（debug 产物 + flycheck + 旧安装包） |
bun run seed:dev |
用生产数据目录播种 dev 数据目录 |
bun run bump |
统一改版本号：根 Cargo.toml(真源) + package.json + ui + Cargo.lock，可选 --tag 提交并打 tag |
bun run help |
打印脚本目录（--check 校验登记 / --readme 生成 README 表） |
docs/README.md— documentation indexdocs/getting-started/— installation and first rundocs/guides/— user & operator guides (companions, channels, AutoWork, knowledge, computer/browser use, terminal, remote API, …)docs/architecture/— technical architecturedocs/reference/— configuration, API overview, FAQ, troubleshooting
Docs are bilingual: every page has an English *.md and a Simplified-Chinese *.zh.md sibling.
OpenHub is pre-1.0 and built part-time, so there's a lot still in flight. On the horizon: prebuilt installers, inbound issue-tracker / requirement sources, more knowledge connectors (Feishu, and beyond), official desktop binaries — plus a few surprises we're genuinely excited about. Stay tuned. ✨
OpenHub very much needs your help to grow — code contributions, community building, and evangelism are all hugely welcome. If you have passion for this project, please reach out and build the OpenHub ecosystem with us.
- Read
CONTRIBUTING.mdto get set up and learn the check ladder. Simplified Chinese:CONTRIBUTING.zh-CN.md. - Be excellent to each other — see
CODE_OF_CONDUCT.md. - Found a vulnerability? Follow
SECURITY.md. - Browse open issues for a place to start.
This is a part-time effort with limited bandwidth, and many delightful features are still on the way. If this resonates with you, join in any way you like — a line of code, a suggestion, a reshare all mean a lot.
OpenHub is completely open source, with nothing held back. Individuals and enterprises are free to build on it and use it commercially.
- Forks & commercial use are welcome. They're also at your own risk — the author and contributors assume no liability for downstream use. Apache-2.0 requires no permission from us.
- A friendly heads-up is appreciated, not required. If you fork or commercialize OpenHub, we'd love a note — not as a license condition, simply because knowing the project is valued is the kind of recognition that keeps it going.
- Some features were intentionally left out of the open-source release to keep the local-data promise airtight — without the people and funding to guarantee everyone's data security, removing them was the responsible choice. As time and resources allow, we hope to bring more of them to you.
Thank you for being here. 🙏
Projects and products we appreciate:
| Product | What it does |
|---|---|
| Saytive | Be Creative, Be Saytive. A voice input method for creative workers, using strong models and thoughtful product design to sense your work context and deliver fast, accurate, scene-aware transcription. |
| Fast | Search, one tap away. Type, click, and jump straight to search results across RED, Douyin, Meituan, and dozens of mainstream apps. No feed distraction, just search. |
| AionUi | AionUi ships with a complete AI agent engine. Unlike tools that require separate CLI-agent installs, AionUi works the moment you install it. |
We'd love to hear from you. The fastest way to reach us is GitHub; the social channels below are all official.
| Channel | Where |
|---|---|
| 🌐 Website | www.openhub.dev |
| 🐙 GitHub | openhub/openhub-client · Issues · Releases |
| ✉️ Email | hello@openhub.dev (provisional — being finalized) |
| 📕 小红书 / RED | OpenHub |
| 📺 Bilibili | OpenHub · demo video |
| 🎵 抖音 / Douyin | OpenHub |
| @OpenHub-o2y · demo video | |
| 𝕏 X (Twitter) | @colir0 |
| 🎬 TikTok | @colir0luo |
Join the chat groups — scan to join:
 WeChat group / 微信群 |
 QQ group / QQ 群 |
Apache-2.0 © 2025–2026 OpenHub.
See NOTICE for third-party attributions.