AI工作流分析
AI工作流分析 是 AI Skill Hub 本期精选Agent工作流之一。综合评分 7.5 分，整体质量较高。我们推荐使用将其纳入你的 AI 工具库，帮助提升工作效率。
📚 深度解析
AI工作流分析 工作流的设计遵循"最小配置，最大复用"原则：核心逻辑已经封装好，用户只需配置自己的 API Key 和业务参数即可快速上手。工作流内置错误处理和重试机制，在网络波动或 API 限速等情况下仍能稳定运行，适合作为生产环境的自动化基础设施。
在实际部署时，建议先在测试环境中运行 3-5 次，验证各个环节的输出结果符合预期，再部署到生产环境。AI Skill Hub 评分 7.5 分，是同类 Agent 工作流中的精选推荐。
📋 工具概览
AI工作流分析 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
📖 中文文档
AI工作流分析 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
- 可视化 Agent 工作流编排，无需编写复杂代码
- 支持多步骤自动化任务链，实现全流程无人值守
- 与外部 API、数据库和第三方服务无缝集成
- 内置错误处理与自动重试机制，保障稳定运行
- 提供可复用的自动化模板，快速在同类场景部署
- 自动化日常重复性工作，将精力集中于创造性任务
- 构建数据采集 → 处理 → 输出的完整自动化管线
- 实现跨平台、跨系统的数据流转和业务协同
# 方式一：npm 全局安装 npm install -g clawd-insights # 方式二：npx 直接运行（无需安装） npx clawd-insights --help # 方式三：项目依赖安装 npm install clawd-insights # 方式四：从源码运行 git clone https://github.com/yx0716/clawd-insights cd clawd-insights npm install npm start
- 访问 GitHub 仓库获取工作流文件
- 在对应平台（Dify / Flowise / Make 等）中找到「导入工作流」功能
- 上传工作流文件
- 按照提示配置必要的环境变量和 API Key
- 运行测试确认流程正常后投入使用
# 命令行使用
clawd-insights --help
# 基本用法
clawd-insights [options] <input>
# Node.js 代码中使用
const clawd_insights = require('clawd-insights');
const result = await clawd_insights.run(options);
console.log(result);
# clawd-insights 配置说明 # 查看配置选项 clawd-insights --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量（覆盖配置文件） export CLAWD_INSIGHTS_CONFIG="/path/to/config.yml"
简介
Features
| Capability | What you get |
|---|---|
| **Timeline view** | Visualize every session by date / project / agent / duration — at a glance, see when you worked, on what, and for how long |
| **Local history scan** | Reads ~/.claude/projects/, ~/.codex/sessions/, ~/.cursor/projects/ directly. No upload, no telemetry |
| **AI session review** | Per-session summary from the *user's* point of view: what you were trying to do, what you walked away with, key topics, time breakdown |
| **Flexible backends** | Local claude CLI, local codex CLI, or fall back to a configured API provider / Ollama — your choice, your keys. Add any number of **custom OpenAI-compatible endpoints** (Zhipu AI, DeepSeek, OpenRouter…) for cheap session analysis |
| **Batch pre-analysis** | Pre-compute summaries for recent sessions and reuse provider-aware cached results |
| **Cost tracking** | See token usage and cost per analysis run |
| **Quick access** | Open from the tray menu, the right-click menu on the desktop pet, or a global shortcut |
Desktop pet capabilities (synced from upstream)
Beyond the analytics layer, this fork now tracks the full desktop-pet feature set from clawd-on-desk:
- Broader agent support — pixel-art reactions for Claude Code, Codex CLI, Copilot CLI, Gemini CLI, Cursor Agent, Kiro CLI, opencode, plus the newly merged Antigravity CLI, CodeBuddy, Hermes, Kimi CLI, openclaw, and Pi.
- WSL & remote development (focus of this sync) — Codex officially supports WSL2, and Clawd integrates through Codex's official hooks (with JSONL polling as a fallback). For an agent running on a remote box or inside WSL's separate Linux home, Remote SSH deploys the hooks over an SSH tunnel so the pet reacts to those sessions too. The remote side must use a POSIX shell — Git Bash or WSL
bash, not Windowscmd.exe. See docs/guides/codex-wsl-clarification.md and docs/guides/setup-guide.md. - GUI settings panel — a full settings window (agents, themes, shortcuts, Remote SSH, Telegram approval) replaces hand-editing config files.
- Theming — swap the crab for alternate characters (e.g. the Cloudling theme) or build your own with
npm run create-theme. - Telegram approval — approve or deny permission requests remotely from your phone.
- Doctor diagnostics — a built-in health check that verifies hook installation and per-agent integration status.
- Session HUD & Quick Commands — optional on-screen session-state labels and a quick-command surface.
- The classics — permission bubbles, mini mode, click reactions, eye-tracking, sleep sequences, and multi-monitor support all carried over.
Platform notes (Windows terminal focus, macOS focus, known limitations) live under docs/guides/.
Quick Install
git clone https://github.com/yx0716/clawd-insights.git && cd clawd-insights && npm install && npm startA small crab appears on your desktop — on macOS, right-click it to open the Analytics Dashboard. For provider setup and analysis workflows, see Getting Started below.
Getting Started
1. Install and run
git clone https://github.com/yx0716/clawd-insights.git && cd clawd-insights && npm install && npm startOnce it launches, a small crab (the default theme) appears on your desktop — that's the Clawd pet, and every entry point to the dashboard goes through it.
Usage examples
① Open right-click → Dashboard |
② Pick a provider Local CLI / API / Ollama |
③ Tweak settings gear ⚙ → AI Provider |
④ Run analysis batch or per-session |
3. Configure an AI Provider for session summaries
The timeline works out of the box. But to make the dashboard automatically generate a recap summary for each session, you need to point it at something that can call a large language model — an AI Provider (the analysis backend). There are three options:
| Provider type | What it is | Setup | Best for |
|---|---|---|---|
| **Local CLI** *(recommended)* | Reuses the claude (Claude Code) or codex CLI you already have installed. Uses your existing subscription, no extra API charges. | **Nothing — auto-detected** | Anyone already using Claude Code or Codex — zero overhead |
| **API key** | An API key from Anthropic, OpenAI, or another provider. Pay-per-token. | Paste your key into the dashboard settings | Users without a local CLI who don't mind a small token cost |
| **Ollama** | A locally-hosted open model server (e.g. Ollama). | Point the dashboard at your Ollama endpoint | Fully offline, never sends data to the cloud |
💡 Strong recommendation: if you already have Claude Code or Codex CLI installed, do nothing — the dashboard auto-detects them and reuses your existing subscription quota. Cheapest and easiest path.
If you don't want to configure a provider right now, click Skip on the startup screen. You can always set it up later in the settings.
<p align="center"> <img src="assets/screen-shot-select-AI-provider.gif" width="720" alt="Selecting and configuring an AI Provider in action"> </p>
4. Where to configure / change the provider later
If you skipped step 3, or you want to switch providers later, you can adjust it via AI Provider Settings at any time:
Open the Analytics Dashboard → click the gear icon ⚙ in the top-right → AI Provider Settings dialog appears.
<p align="center"> <img src="assets/screenshot-ai-provider-settings.gif" width="720" alt="AI Provider Settings dialog"> </p>
The dialog has two sections:
- LOCAL CLI DETECTION — shows whether the dashboard found
claudeandcodexon your machine. Green dot = found (with version + path); red dot = missing. If you see green dots, everything is working — proceed to the next step. - API PROVIDER (FALLBACK) — if no local CLI is installed, you can use an API key for AI session analysis (Claude / OpenAI / Ollama / …) — just paste the key and you're set.
- CUSTOM ANALYSIS PROVIDERS — add any number of OpenAI-compatible endpoints (Zhipu AI, DeepSeek, OpenRouter, university APIs, etc.) as dedicated analysis backends. See below.
Tip: if yourclaudeorcodexwas installed via NVM, fnm, or Volta, auto-detection may miss it. Runwhich claudeorwhich codexin your terminal and paste the output into the Claude binary path / Codex binary path override field.
FAQ
Q: Does the dashboard need internet? Scanning and the timeline are fully offline. Whether AI summaries need internet depends on which provider you pick: Local CLI uses whatever network stack Claude Code / Codex normally use; Ollama is fully offline; API key mode talks to the cloud.
Q: Are my conversations uploaded anywhere? No. Clawd Insights collects zero telemetry. The provider step is your CLI or your API key calling the model you chose directly — no third-party server in the middle.
Q: I don't have Claude Code or Codex. Can I still use it? Yes. You can use the timeline view alone (completely free, no LLM required), or paste an Anthropic / OpenAI API key into AI Provider Settings to enable the cloud path.
高质量的AI工作流分析工具
- 构建多智能体协作系统的 Agent 开发者
- 本地部署优先选 GGUF 量化模型，节省显存并保持响应速度
- Agent 任务先做 dry-run 验证工具调用链，再开启自主执行
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
- 显存不足直接 OOM — 优先降低 context 或换更小的量化模型
- CLI：直接 npm install -g / pip install，命令行调用
- 本地部署：CPU 8GB 起，GPU 推荐 16GB+ 显存
- 云端托管：可放在 Vercel / Railway / Fly.io 等 PaaS 平台
⚡ 核心功能
- 可视化 Agent 工作流编排，无需编写复杂代码
- 支持多步骤自动化任务链，实现全流程无人值守
- 与外部 API、数据库和第三方服务无缝集成
- 内置错误处理与自动重试机制，保障稳定运行
- 提供可复用的自动化模板，快速在同类场景部署
- 构建多智能体协作系统的 Agent 开发者
- 本地部署优先选 GGUF 量化模型，节省显存并保持响应速度
- Agent 任务先做 dry-run 验证工具调用链，再开启自主执行
- API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
- 显存不足直接 OOM — 优先降低 context 或换更小的量化模型
👥 适合人群
🎯 使用场景
- 自动化日常重复性工作，将精力集中于创造性任务
- 构建数据采集 → 处理 → 输出的完整自动化管线
- 实现跨平台、跨系统的数据流转和业务协同
⚖️ 优点与不足
- +大幅减少重复性人工操作
- +可视化流程，清晰直观
- +可扩展性强，支持复杂场景
- −初始配置和调试需投入一定时间
- −强依赖外部服务的稳定性
- −复杂场景需具备一定技术基础
该工具使用 AGPL-3.0 协议，商用场景请仔细阅读协议条款，必要时咨询法律意见。
AI Skill Hub 为第三方内容聚合平台，本页面信息基于公开数据整理，不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后，再部署至生产环境，并做好必要的安全评估。
⚠️ AGPL 3.0 — 最严格的 Copyleft，网络服务端使用也需开源，SaaS 使用受限。
🔗 相关工具推荐
❓ 常见问题 FAQ
经综合评估，AI工作流分析 在Agent工作流赛道中表现稳健，质量良好。如果你已有明确的使用需求，可以直接上手体验；如果还在评估阶段，建议对比同类工具后再做决策。
| 原始名称 | clawd-insights |
| 原始描述 | 开源AI工作流：Local-first analytics dashboard + desktop pet for your Claude Code, Codex CLI an。⭐36 · JavaScript |
| Topics | ai-agentsanalyticsclaude-codecodexdesktop-pet |
| GitHub | https://github.com/yx0716/clawd-insights |
| License | AGPL-3.0 |
| 语言 | JavaScript |
收录时间：2026-05-26 · 更新时间：2026-05-30 · License：AGPL-3.0 · AI Skill Hub 不对第三方内容的准确性作法律背书。
🤖 交给 Agent 安装 · AI工作流分析
选择 Agent 类型，复制安装指令后粘贴到对应客户端