# Agentic AI

# Agentic Tools

#### Agent Development

- Microsoft: [10 Lessons teaching everything you need to know to start building AI Agents](https://github.com/microsoft/ai-agents-for-beginners)
- Kaggle: [5-Day AI Agents Intensive Course with Google | Kaggle](https://www.kaggle.com/learn-guide/5-day-agents)
- Google: [Agent Development Kit](https://google.github.io/adk-docs/)

#### Top AI Agents Platforms

No code 與 Low code AI 應用開發平台

##### Self-Hosted

1. [n8n](https://n8n.io/)
2. [CrewAI](https://www.crewai.com/open-source)
3. [Dify](https://dify.ai/zh)
4. [LangFlow](https://www.langflow.org/) - 例如基於 RAG 的 AI 整合應用

##### Cloud-Based

- [Coze](https://www.coze.com/)
- [Zapier](https://zapier.com/)
- [Make](https://www.make.com/en)

#### n8n

- [https://n8n.io/](https://n8n.io/)
- Doc: [https://docs.n8n.io/hosting/](https://docs.n8n.io/hosting/)
- [Automate Your Home Lab with n8n Workflow Automation and AI - Virtualization Howto](https://www.virtualizationhowto.com/2025/07/automate-your-home-lab-with-n8n-workflow-automation-and-ai/)
- [【n8n 中文教學】新手入門：介面功能說明、Webhook 和自動化工作流應用](https://raymondhouch.com/lifehacker/digital-workflow/n8n-beginner-guide/)
- [「超詳細教學」n8n AI 實作0基礎入門到進階 — (AI Agent | LLM | RAG | Webhook| AI 自動生成研究報告) - YouTube](https://www.youtube.com/watch?v=vvqhzbp4J5A)

##### Community Nodes

- [Install and manage community nodes | n8n Docs](https://docs.n8n.io/integrations/community-nodes/installation/)

PDF to Image

- [n8n-nodes-pdf2image](https://github.com/guziakas/n8n-pdf2image)
- [n8n-nodes-pdfconvert](https://ncnodes.com/package/n8n-nodes-pdfconvert)

##### n8n Installation

On a laptop (small level)

- YOUR\_TIMEZONE: `Asia/Taipei` or `America/Chicago` or `America/New_York`

```bash
docker volume create n8n_data

docker run -it --rm \
 --name n8n \
 -p 5678:5678 \
 -e GENERIC_TIMEZONE="<YOUR_TIMEZONE>" \
 -e TZ="<YOUR_TIMEZONE>" \
 -e N8N_ENFORCE_SETTINGS_FILE_PERMISSIONS=true \
 -e N8N_RUNNERS_ENABLED=true \
 -v n8n_data:/home/node/.n8n \
 docker.n8n.io/n8nio/n8n
```

##### Quick Start Examples

Basic Data Access

```javascript
{{ $json.fieldName }}                    // Get field from current item
{{ $('Node Name').item.json.field }}     // Data from specific node
{{ $json.items?.[0] }}                   // Safe array access
```

Date Functions

```javascript
{{ $now.format('YYYY-MM-DD') }}          // 2024-01-15
{{ $now.plus({days: 7}) }}               // 7 days from now
{{ $now.diff('2024-01-01', 'days') }}    // Days between dates
```

String Manipulation

```javascript
{{ $json.name.toLowerCase() }}           // convert to lowercase
{{ $json.email.split('@')[0] }}          // Username from email
{{ $json.text.slice(0, 50) }}            // First 50 characters
```

Array Processing

```javascript
{{ $json.items.length }}                 // Count array items
{{ $json.items.filter(item => item.active) }}   // Filter items
{{ $json.items.map(item => item.name) }}        // Extract field from all items
```

##### 跨 Nodes 傳 binary 內容

- [Analyze any Video, Image or PDF with Gemini and n8n (Step-by-Step) - YouTube](https://www.youtube.com/watch?v=OWJKGkPEGSw) (Gemini API - Http Request)

新增 Code node

```javascript
return{
  json: {},
  binary: $('On Form submission').item.binary
}
```

##### Gmail Trigger

- [I Made an Email Auto-Responder to Conquer my Writer’s Block | by Nicolas CHOURROUT | Medium](https://medium.com/@nchourrout/i-made-an-email-auto-responder-to-conquer-my-writers-block-aa2b91db6741)

#### MCP

- [Building AI Agents with Google ADK, FastAPI, and MCP - DEV Community](https://dev.to/timtech4u/building-ai-agents-with-google-adk-fastapi-and-mcp-26h7)
- [Awesome MCP Servers](https://mcpservers.org/zh-TW/)

#### No Code Tools

- [Airtable](https://www.airtable.com/) - 雲端資料庫 
    - [【中文介紹】什麼是 Airtable？最強大的雲端資料庫教學和應用範例](https://raymondhouch.com/lifehacker/digital-workflow/airtable/)
- [CodeGraph](https://github.com/colbymchenry/codegraph) - 將本地程式專案轉成 AI 看得懂的知識圖譜
- [GitNexus](https://github.com/abhigyanpatwari/GitNexus) - 將 GitHub 專案轉成知識圖譜

#### Tool Calling

- [Mastering LLM Tool Calling: The Complete Framework for Connecting Models to the Real World - MachineLearningMastery.com](https://machinelearningmastery.com/mastering-llm-tool-calling-the-complete-framework-for-connecting-models-to-the-real-world/)

#### Alternatives to OpenClaw

##### Hermes Agent

The self-improving AI agent built by Nous Research. It's the only agent with a built-in learning loop — it creates skills from experience, improves them during use, nudges itself to persist knowledge, searches its own past conversations, and builds a deepening model of who you are across sessions.

- [https://github.com/nousresearch/hermes-agent](https://github.com/nousresearch/hermes-agent)

##### DeerFlow

DeerFlow (Deep Exploration and Efficient Research Flow) is an open-source super agent harness that orchestrates sub-agents, memory, and sandboxes to do almost anything — powered by extensible skills.

- [https://github.com/bytedance/deer-flow](https://github.com/bytedance/deer-flow)

# OpenClaw

#### Introduction

[OpenClaw](https://openclaw.ai/) 是一款開源的個人 AI 數位助理，讓你在各種通訊平台上擁有專屬的智慧管家。無論是自動回覆訊息、整理資訊、串接工作流程，OpenClaw 都能幫你搞定。

- 中文手冊：[OpenClaw：在 WhatsApp 與 Telegram 輕鬆使用你的 AI Agent | OpenClaw](https://open-claw.bot/docs/zh-tw/)
- 李宏毅: [解剖小龍蝦 — 以 OpenClaw 為例介紹 AI Agent 的運作原理 - YouTube](https://www.youtube.com/watch?v=2rcJdFuNbZQ)
- [為什麼我拖了一個多月才開始使用OpenClaw？ （附接入飛書完整詳細教程，新手必看）](https://codelove.tw/@tony/post/am2Zoq)
- [OpenClaw 教學：26 個 Tools + 53 個 Skills 完整指南 | WenHao Yu](https://yu-wenhao.com/zh-TW/blog/openclaw-tools-skills-tutorial/)
- [OpenClaw：在 WhatsApp 與 Telegram 輕鬆使用你的 AI Agent | OpenClaw](https://open-claw.bot/docs/zh-tw/)
- [OpenClaw 中文教程 - 个人 AI 智能体开发一站式社区](https://openclawai.cc/)
- [OpenClaw 新手必备！安装实用Skills，模型选择，浏览器自动化等！ – 零度博客](https://www.freedidi.com/23203.html)

<span style="color: rgb(187, 187, 187); font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Oxygen, Ubuntu, Roboto, Cantarell, 'Fira Sans', 'Droid Sans', 'Helvetica Neue', sans-serif; font-size: 1.4em; font-weight: 400;">Resources</span>

- [Awesome OpenClaw Agents](https://github.com/mergisi/awesome-openclaw-agents)
- [ClawHub](https://docs.openclaw.ai/zh-CN/tools/clawhub)
- [OpenClaw CLI](https://skill-deploy-wwjmzwrcj4.vercel.app/)
- [建議模型列表](https://pinchbench.com/)
- [各種專家級 AI 助理](https://github.com/msitarzewski/agency-agents)
- [193 個即插即用的 AI 專家角色](https://github.com/jnMetaCode/agency-agents-zh)

#### Installation

- [【保姆级】OpenClaw 全网最细教学：安装→Skills实战→多Agent协作，1 小时全精通！ | 小林的博客-AI学长](https://ai.linbintalk.com/article/ytopenclaw)
- YT: [【保姆级】OpenClaw 全网最细教学：安装→Skills实战→多Agent协作，1 小时全精通！ - YouTube](https://www.youtube.com/watch?v=2ZZCyHzo9as)

##### With Docker

- [OpenClaw Docker 部署完全指南：从零搭建你的私有 AI Agent](https://ofox.ai/zh/blog/openclaw-docker-deployment-guide-2026/)
- [Self-hosting OpenClaw with Docker and Tailscale on a $5 VPS](https://heyabhishek.com/blog/self-hosting-openclaw-docker-tailscale/)
- [Docker 部署 OpenClaw：从踩坑到跑通的完整记录 | Yuxu Ge](https://yuxu.ge/blog/2026/2026-02-25-openclaw-docker-deploy-zh.html)

```bash
git clone https://github.com/openclaw/openclaw
cd openclaw
export OPENCLAW_IMAGE="ghcr.io/openclaw/openclaw:latest"
./scripts/docker/setup.sh
```

##### Debian 13

Run as root

```bash
# Install Node.js
# Run as root
curl -fsSL https://deb.nodesource.com/setup_24.x | bash -
apt install nodejs
```

Run as non-root

```bash
# Switch to non-root account
# NOTE: Node.js 22.14 or newer is required for openclaw
su - <your-username>
node -v
npm --version

mkdir -p "$HOME/.npm-global"
npm config set prefix "$HOME/.npm-global"
npm prefix -g
export PATH="$HOME/.npm-global/bin:$PATH"
```

Edit: ~/.bashrc

```shell
export PATH="$HOME/.npm-global/bin:$PATH"
```

Install openclaw

```bash
npm install -g openclaw@latest
openclaw onboard --install-daemon
```

#### Post-Install

##### 訂製 AI 助理

使用 3 個檔案量身訂製助理

- SOUL.md — 身份、記憶、溝通風格、關鍵規則
- AGENTS.md — 核心使命、技術交付物、工作流程
- IDENTITY.md — 名稱與簡介

##### Timezone

```bash
openclaw config set agents.defaults.userTimezone "Asia/Taipei"
```

#### CLI Commands

##### Check Service

```bash
docker compose run --rm openclaw-cli status
docker compose run --rm openclaw-cli gateway status
```

##### Pairing required

Approve the device

```bash
docker compose run --rm openclaw-cli devices list
docker compose run --rm openclaw-cli devices approve <request-id>
```

##### Model Configuration

```bash
# 主要模型設置
openclaw models list
openclaw models status
openclaw models set <provider/model>
openclaw models set-image <provider/model>

# 備用模型設置
openclaw models fallbacks list
openclaw models fallbacks add <provider/model>
openclaw models fallbacks remove <provider/model>
openclaw models fallbacks clear
```

#### Telegram

##### Configuration

Create a Bot from Telegram

1. Open Telegram
2. Chat with `@BotFather`
3. Run `/newbot` , follow prompts to create your bot, and save the token. 
    - Bot Name: &lt;whatever&gt;
    - Bot Username: &lt;unique-name and must end in 'bot'&gt;

Go to OpenClaw CLI

```bash
docker compose run --rm openclaw-cli channels add --channel telegram --token "<bot-token>"
```

Go to Telegram

1. Open you-bot
2. Send "Hi"
3. Getting a pairing code "XXXXX"

Go to OpenClaw CLI

```bash
docker compose run --rm openclaw-cli pairing list telegram
docker compose run --rm openclaw-cli pairing approve telegram XXXXX
```

Done

##### Bot Commands

- `/btw` : 題外話，在不打斷目前對話（主任務），可以臨時問一個小問題，其內容不會進入主對話的上下文。
- `/subagent` : 子代理，在不打斷目前對話（主任務），呼叫另一個代理執行其他任務，其任務結果會納入主任務的上下文。
- `/new` : 新對話，建立新的對話（任務），全新的上下文，可節省 Token 用量。

#### Gateway

##### Local only (default)

- 遠端存取可以使用 SSH Port Forwarding

`~/.openclaw/openclaw.json` :

```json
  "gateway": {
    "port": 18789,
    "mode": "local",
    "bind": "loopback",
    "controlUi": {
      "allowedOrigins": [
        "http://localhost:18789",
        "http://127.0.0.1:18789"
      ]
    },
    "auth": {
      "mode": "token",
      "token": "YOUR-OPENCLAW-TOKEN"
    },
```

##### Tailscale

<p class="callout info">openclaw.json 的 tailscale 功能不適用在 Docker 環境。</p>

1. Tailscale Console: Enable **HTTPS Certicficates**
2. OpenClaw Host: Run 
    1. `sudo tailscale serve --bg --https=443 127.0.0.1:18789`
    2. `tailscale serve status`
3. Change `~/.openclaw/openclaw.json` as follows
4. Connect to tailnet, and then visit https://your-openclaw-device.tailnet-domain/?token=YOUR-TOKEN , such as *https://dockers-vm.tailcb58c9.ts.net/* ?token=YOUR-TOKEN

`~/.openclaw/openclaw.json` :

- mode: local
- bind: loopback
- trustedProxies: \["127.0.0.1", "::1"\]
- allowedOrigins: https://yourdevice.tailnet-domain

```json
  "gateway": {
    "port": 18789,
    "mode": "local",
    "bind": "loopback",
    "trustedProxies": ["127.0.0.1", "::1"],
    "controlUi": {
      "allowedOrigins": [
        "http://localhost:18789",
        "http://127.0.0.1:18789",
        "https://yourdevice.tailnet-domain"
      ]
    },
    "auth": {
      "mode": "token",
      "token": "YOUR-OPENCLAW-TOKEN"
    },
```

openclaw.json 的 tailscale 模式比較（不適用在 Docker 環境）

<table border="1" id="bkmrk-%E6%A8%A1%E5%BC%8F-%E5%AD%98%E5%8F%96%E7%AF%84%E5%9C%8D-https-%E8%AA%8D%E8%AD%89-%E9%81%A9%E7%94%A8%E5%A0%B4" style="border-collapse: collapse; width: 100%;"><colgroup><col style="width: 20%;"></col><col style="width: 20%;"></col><col style="width: 20%;"></col><col style="width: 20%;"></col><col style="width: 20%;"></col></colgroup><tbody><tr><td>模式</td><td>存取範圍</td><td>HTTPS</td><td>認證</td><td>適用場景</td></tr><tr><td>serve</td><td>限 tailnet</td><td>自動</td><td>Tailscale identity headers/Token/Password</td><td>個人使用</td></tr><tr><td>funnel</td><td>公眾網路</td><td>自動</td><td>Password</td><td>Webhook</td></tr><tr><td>bind: "tailnet"</td><td>限 tailnet</td><td>無</td><td>Token/Password</td><td>低延遲</td></tr><tr><td>off</td><td>限 localhost</td><td>無</td><td>Token</td><td>SSH Port Forward</td></tr></tbody></table>

#### Skill

##### 3rd Party

- [Architecture Diagram Generator](https://github.com/Cocoon-AI/architecture-diagram-generator): 繪製架構圖並輸出 HTML 格式。
- [Khazix Skills](https://github.com/KKKKhazix/khazix-skills): 橫縱分析法深度研究。

#### FAQ

##### Context limit exceeded

> Context limit exceeded. I've reset our conversation to start fresh - please try again.
> 
> To prevent this, increase your compaction buffer by setting agents.defaults.compaction.reserveTokensFloor to 20000 or higher in your config.

Cause: 目前使用模型的 Context Window 與 壓縮緩衝參數不匹配。

Solution: 以 Gemma4-31B-it 為例

Edit: ~/.openclaw/openclaw.json

- `contextWindow` : 128000
- `reserveTokensFloor` : 40000

```json
  "models": {
    "mode": "merge",
    "providers": {
      "custom-cpamc": {
        "baseUrl": "http://192.168.31.89:8317/v1",
        "api": "openai-completions",
        "apiKey": "sk-cvgaT1Z3EhJRBB5pu",
        "models": [
          {
            "id": "gemma-4-31b-it",
            "name": "gemma-4-31b-it (Custom Provider)",
            "contextWindow": 128000,
            "maxTokens": 4096,
            "input": [
              "text"
            ],
```

```json
  "agents": {
    "defaults": {
      "workspace": "/home/alang/.openclaw/workspace",
      "model": {
        "primary": "custom-cpamc/gemma-4-31b-it"
      },
      "models": {
        "custom-cpamc/gemma-4-31b-it": {}
      },
      "compaction": {
        "reserveTokensFloor": 40000
      }
    }
  },
```

或者使用 CLI

```bash
openclaw config set agents.defaults.compaction.reserveTokensFloor 40000
```

##### LLM idle timeout

> The model did not produce a response before the LLM idle timeout. Please try again, or increase agents.defaults.llm.idleTimeoutSeconds in your config (set to 0 to disable).

Solution:

```bash
openclaw config set agents.defaults.timeoutSeconds 600
openclaw config set agents.defaults.llm.idleTimeoutSeconds 600
```

##### gateway connect failed

> gateway connect failed: GatewayClientRequestError: pairing required

檢查有無 pending devices

```bash
openclaw devices list
openclaw devices list --json
```

# Hermes Agent

由 Nous Research 打造的自我進化 AI Agent。唯一內建學習循環的 Agent — 它能從經驗中建立技能，在使用過程中不斷優化，主動提醒自己持久化知識，並在多次會話中逐步加深對你的理解。

- [Hermes Agent — The Agent That Grows With You | Nous Research](https://hermes-agent.nousresearch.com/)
- [https://github.com/nousresearch/hermes-agent](https://github.com/nousresearch/hermes-agent)

Tutorials

- [Hermes Agent 文档 | Hermes Agent 中文文档](https://hermes-doc.aigc.green/)
- [Hermes Agent 部署 UI + 本地模型 Gemma 4，对接微信（完全免费，无需 Tokens） - 零度博客](https://www.freedidi.com/23749.html)

Resources

- [Skills Hub | Hermes Agent](https://hermes-agent.nousresearch.com/docs/skills)
- [Awesome Hermes Agent](https://github.com/0xNyk/awesome-hermes-agent)
- [Hermes Atlas](https://hermesatlas.com/) - 各種工具、技能、外掛及整合項目
- [Hermes HUD](https://github.com/joeynyc/hermes-hudui) - 非官方 Hermes Dashboard 
    - [8萬顆星的Hermes Agent少了它根本不能用！3分鐘打造AI中控台 - YouTube](https://www.youtube.com/watch?v=78YQMfgPlwA)

#### Installation

Debian/Ubuntu

```bash
# Install
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

# Configure
hermes setup
```

##### Post-Install

訂製 AI 助理

- `~/.hermes/SOUL.md`
- `~/.hermes/memories/USER.md`

#### Web UI

##### Hermes-WebUI

- [https://github.com/nesquena/hermes-webui](https://github.com/nesquena/hermes-webui)

```bash
git clone https://github.com/nesquena/hermes-webui.git
cd hermes-webui
./start.sh
```

##### hermes-web-ui

- [https://github.com/EKKOLearnAI/hermes-web-ui](https://github.com/EKKOLearnAI/hermes-web-ui)

# Agent Skills

#### Resources

- [skills.sh](https://skills.sh/)
- [Skill 手册 | 01MVP Docs | 01MVP Starter Kit](https://www.01mvp.com/docs/skills)
- [antigravity-awesome-skills](https://github.com/sickn33/antigravity-awesome-skills)
- [Asgard Skills](https://github.com/asgard-ai-platform/skills) - 開源的 293 個 coding agent skills 知識庫，分成 22 個主題類別。每個 skill 都是獨立的 Markdown 檔案（SKILL.md），遵循 Claude Agent Skills 規範，部分附帶純 Python 腳本做確定性計算。

#### 語音生成

- edge-tts

#### 影片生成

- hyperframe

#### 網頁搜尋

- tavily-search
- brave-search (badlogic)

#### PDF

- pdf (openai)
- [LiteParse](https://github.com/run-llama/liteparse) - A fast, helpful, and open-source document parser
- [pdftk-server](https://www.skills.sh/github/awesome-copilot/pdftk-server) - 合併、分割、加密等各種 PDF 進階操作

#### Web Scraper

##### - Crawlee

A web scraping and browser automation library.

- [crawlee for Python](https://crawlee.dev/python)
    - GitHub: [https://github.com/apify/crawlee-python](https://github.com/apify/crawlee-python)
- [crawlee for Node.js](https://crawlee.dev/)
    - GitHub: [https://github.com/apify/crawlee](https://github.com/apify/crawlee)

##### - ScrapeGraphAI

[ScrapeGraphAI](https://scrapegraphai.com/) is a open-source web scraping python library designed to usher in a new era of scraping tools.

- [Indices and tables — ScrapeGraphAI documentation](https://scrapegraph-ai.readthedocs.io/en/latest/)
- GitHub: [https://github.com/ScrapeGraphAI/Scrapegraph-ai](https://github.com/ScrapeGraphAI/Scrapegraph-ai)
- Video: [Scrape Any Website using llama3+Ollama+ScrapeGraphAI | Fully Local + Free #ai #llm - YouTube](https://www.youtube.com/watch?v=2BTI3KIiGHU)

##### - Crew AI

[Crew AI](https://crewai.net/) is a collaborative working system designed to enable various artificial intelligence agents to work together as a team, efficiently accomplishing complex tasks. Each agent has a specific role, resembling a team composed of researchers, writers, and planners.

- GitHub: [https://github.com/joaomdmoura/crewAI](https://github.com/joaomdmoura/crewAI)
- 範例：使 AI 自動爬文並生成筆記 
    - Video: [如何搭建一套Agent系统](https://www.youtube.com/watch?v=vEC8NLkc4rk)
    - GitHub: [Python 程式碼](https://github.com/business24ai/crewai-obsidian)
- [Crew AI — your own minions. How I Made AI Assistants Do My Work For… | by Csakash | Medium](https://medium.com/@csakash03/crew-ai-you-own-minions-9b8596ce3da3)

##### - Scraperr

Self-hosted webscraper.

- [Welcome to the Scraperr Docs | Scraperr Docs](https://scraperr-docs.pages.dev/)
- GitHub: [https://github.com/jaypyles/Scraperr](https://github.com/jaypyles/Scraperr)

##### - Crawl4AI

Open-source LLM Friendly Web Crawler &amp; Scraper.

- GitHub: [https://github.com/unclecode/crawl4ai](https://github.com/unclecode/crawl4ai)

##### - Scrapling

Effortless Web Scraping for the Modern Web

- [https://github.com/D4Vinci/Scrapling](https://github.com/D4Vinci/Scrapling)
- [当下最强免费爬虫工具！Scrapling保姆级教程（2026最新爬虫神器） - YouTube](https://www.youtube.com/watch?v=fMpdgrZmIgg)

##### - Obscura

The open-source headless browser for AI agents and web scraping.

- [https://github.com/h4ckf0r0day/obscura](https://github.com/h4ckf0r0day/obscura)

# Pi Agent

Pi 是一款極簡的終端編碼工具。讓 Pi 適應您的工作流程，而非讓您去適應它。

官網：[Pi Coding Agent](https://pi.dev/)

#### Tutorials

- [Pi Agent：比Claude Code和Codex更适合普通人的AI工具 - YouTube](https://www.youtube.com/watch?v=mNsqiALIoRI)
- [Setting Up and Using the Pi Coding Agent | DeepakNess](https://deepakness.com/blog/pi-agent-setup/)