autoMate
A personal AI assistant server that provides MCP tools for notes, files, reminders, memory, shell, browser, and 30+ SaaS integrations, enabling AI clients to access and manage your data across sessions.
README
autoMate
A smart NAS for AI. Notes · files · reminders · memory · 40+ tools. Plug it into OpenClaw / Claude Desktop / Cursor / Cline as a tool source, or use it standalone via its built-in web chat.
┌─ OpenClaw ─────┐ ┌─ notes / memory ── survives every chat
├─ Claude Desktop├──── MCP ────┐ ├─ files ─────────── your file vault (NAS)
├─ Cursor / Cline├──── HTTP ───┤ ├─ reminders ─────── push to your phone
├─ Kimi / GPT ├──── bridge ─┤ ├─ memory ────────── cross-session facts
└─ your scripts ┘ │ ├─ search.find ───── BM25 over notes+files
▼ │
┌──────────────────┐ ┌─ shell · script · browser
│ autoMate │ ── ┤ desktop · 31 SaaS APIs
│ (your machine) │ └─ your real Chrome (extension)
└────────┬─────────┘
▼
~/.automate/ · SQLite + Fernet
What's the deal
Every AI vendor wants to be your chat tool. Most can call tools. None of them remember anything across vendors, store your files, or ping your phone when something matters.
autoMate is the layer behind the chat: a warehouse you own. Pick a chat client you like (OpenClaw for IM, Claude Desktop for serious work, Cursor for code), point it at autoMate via MCP, and that client can write into your notes, drop files, schedule reminders, search your library, and call your real-world tools — and tomorrow's client can read all of it back the same way.
Two ways to use it
| Mode | Who's the brain | When |
|---|---|---|
| As a tool inside another AI client | Your client (OpenClaw / Claude / Cursor / ...) | Most use cases — IM, coding, research |
| Standalone via autoMate's web chat | autoMate's own agent loop | Quick local queries, no other client |
Same backend. Same data. Pick the entry point.
Connecting to OpenClaw / Claude Desktop / Cursor / Cline (v4.5.7)
After install, open Settings → Connect to AI clients and click "Copy install text". You get a single markdown blob with the URL + token already filled in, and per-client sections for OpenClaw, Claude Desktop, Cursor, Cline, generic MCP, and non-MCP gateways.
Three ways to use it:
- Read it and edit your client's config file by hand.
- Paste it into another AI — "Cursor, here's autoMate, set it up for me." The text is written so an AI reader can find the right section and edit the right config.
- For OpenClaw specifically, paste into your OpenClaw config under
bundle-mcp— the section spells it out.
After the client picks up the new server, it gets all autoMate's
tools (search.find, notes.read, files.list, audio.transcribe,
...) plus a top-level automate tool that runs autoMate's own
agent loop on demand.
See docs/channels.md for details.
Install
| Path | Get | When |
|---|---|---|
pip install automate-hub |
Python package | Have Python, want it small |
| Standalone binary (Win / macOS / Linux) | Releases | No Python, double-click |
| Docker | docker run -p 8765:8765 ghcr.io/yuruotong1/automate:latest |
Headless box / NAS |
| Browser extension | extension/ |
Drive your real Chrome |
| Android APK | Releases | Optional viewer for the hub |
After install:
automate # double-click on Windows/macOS does the same thing
Browser opens to http://127.0.0.1:8765. The wizard walks you through
picking a model, pasting a key, and (optionally) wiring up an AI client.
What's in the box
Personal data
notes.*— markdown documents with tags, search, pinningfiles.*— content-addressed blob vault, deduplication, configurable storage path (point it at an external SSD or NAS mount)search.find— Coze-style hybrid retrieval (SQLite FTS5 BM25) across notes + files in one callreminders.*— scheduler thread fires Web Push to your PWAmemory.*— long-term key-value facts that any AI can read/writeaudio.transcribe— voice → text via Tencent ASR / OpenAI Whisper, with custom vocabulary mined from your notes (Pro tier)
Local executors
shell.*,script.*(Python/Bash/Node),desktop.*(pyautogui)browser.*(Playwright, fresh Chromium tab)bx.*(your real browser via the Chrome extension)
SaaS integrations — 31 platforms: GitHub · GitLab · Gitee · Notion · Slack · Linear · Jira · Confluence · Trello · Asana · Monday.com · HubSpot · Airtable · Stripe · Shopify · Telegram · Discord · MS Teams · Zoom · Twitter/X · SendGrid · Mailchimp · Twilio · Sentry · 飞书 · 钉钉 · 企业微信 · 微信公众号 · 微博 · 语雀 · 高德地图.
LLM providers — 25 in the catalog: OpenAI · Anthropic · Gemini · xAI Grok · Mistral · Cohere · OpenRouter · Groq · Together · Fireworks · DeepInfra · DeepSeek · Moonshot Kimi · 通义 · 豆包 · GLM · 百川 · Yi · MiniMax · 阶跃 · 混元 · 硅基流动 · Ollama · LM Studio · any OpenAI-compatible.
IM channels (WeChat / Telegram / WhatsApp / ...)
We don't ship per-platform bots ourselves. Run OpenClaw alongside autoMate — it has the official Tencent WeChat plugin for 微信个人助手 (no account-takeover, no ban risk), plus Telegram / WhatsApp / Slack / Discord / Signal / iMessage. autoMate plugs in as its tool source via the MCP setup above. The user talks in their chat platform → OpenClaw routes → OpenClaw's agent calls autoMate when it needs your data.
The legacy bots in automate/bots/ (telegram / wechat_oa / wecom)
are frozen but still ship for backward compatibility.
Project layout
autoMate/
├─ automate/ # the package
│ ├─ server/ # FastAPI app, REST + WS + MCP-over-HTTP
│ │ └─ mcp_bridge.py # FastMCP exposure, mounted at /mcp/
│ ├─ agent/ # NL → tool-call loop
│ ├─ providers/ # LLM provider catalog + clients
│ ├─ tools/ # shell · script · browser · desktop · bx
│ │ # plus notes · files · reminders · memory
│ │ # · search · audio
│ ├─ integrations/ # 31 SaaS connectors
│ ├─ frontend/ # static SPA — PWA installable
│ │ └─ local-store.js # IndexedDB store: notes / memory / files
│ ├─ channels.py # external IM gateway bridge
│ ├─ audio.py # transcription provider abstraction
│ ├─ auth.py # autoMate Cloud session (Pro tier hook)
│ └─ {notes,files,reminders,memory,push}.py
├─ android/ # native Android WebView app (APK)
├─ extension/ # Chrome MV3 extension
├─ docs/{channels,cloud,relay,mobile,sync}.md
├─ Dockerfile
└─ pyproject.toml
Privacy & safety
- Server binds to
127.0.0.1. Network access is opt-in (--host 0.0.0.0). - API keys, OAuth tokens, push subscriptions: encrypted with Fernet,
key at
~/.automate/secret.key(chmod 600). - LLM calls go directly from autoMate to the provider you chose. Nothing else sees them.
- The MCP endpoint at
/mcp/requires a Bearer token. Treat it like a password — anyone with this token can call autoMate's tools, includingshell.exec. Regenerate from Settings → Channels. - autoMate Cloud (paid tier) is opt-in. Without
AUTOMATE_CLOUD_URLset, no data leaves your machine. See docs/cloud.md for the open-client / closed-server boundary.
Status
v4.5.7 — autoMate is now an MCP-over-HTTP tool source for any OpenClaw / Claude Desktop / Cursor / Cline / Cline / Kimi setup. Settings gives you a one-click "Copy install text" for any AI client. Personal infra is filled out: Coze-style retrieval, audio transcription, configurable storage path, in-app auto-update, login hook for the upcoming autoMate Cloud Pro tier.
中文版: README_CN.md
License
MIT.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。