📄 工具详情 ⚙️ 安装教程 📚 使用教程
能力标签
🔌 MCP🤖 Agent🔄 工作流💻 CLI🔗 REST API🧠 Claude🖥 本地 LLM
Claude 代理仪表盘
⚙️
Agent工作流

Claude 代理仪表盘

基于 Python · 无代码搭建完整 AI 自动化流程
英文名：claude-agents-dashboard
⭐ 6 Stars 💻 Python 📄 MIT 🏷 AI 7.5分
7.5AI 综合评分
AI工作流Scrum
⬇ 下载源码 ZIP ⚙️ 配置说明
✦ AI Skill Hub 推荐

Claude 代理仪表盘 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

Claude 代理仪表盘 是一套完整的 AI Agent 自动化工作流方案。随着 AI 能力的不断提升，基于 Agent 的自动化工作流正在成为提升个人和团队效率的核心方式。区别于传统的 RPA 自动化（模拟鼠标键盘操作），AI Agent 工作流通过理解任务意图、动态规划执行路径，能够处理更复杂的非结构化任务。

Claude 代理仪表盘 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

独立的Scrum板，协调Claude代理为项目工作

Claude 代理仪表盘 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。

GitHub Stars
⭐ 6
开发语言
Python
支持平台
Windows / macOS / Linux
维护状态
轻量级项目，按需更新
开源协议
MIT
AI 综合评分
7.5 分
工具类型
Agent工作流
Forks
—

📖 中文文档

以下内容由 AI Skill Hub 根据项目信息自动整理，如需查看完整原始文档请访问底部「原始来源」。

独立的Scrum板，协调Claude代理为项目工作

Claude 代理仪表盘 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。

📌 核心特色
  • 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
🎯 主要使用场景
  • 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
以下安装命令基于项目开发语言和类型自动生成，实际以官方 README 为准。
安装命令
# 方式一：pip 安装（推荐）
pip install claude-agents-dashboard

# 方式二：虚拟环境安装（推荐生产环境）
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install claude-agents-dashboard

# 方式三：从源码安装（获取最新功能）
git clone https://github.com/epatel/claude-agents-dashboard
cd claude-agents-dashboard
pip install -e .

# 验证安装
python -c "import claude_agents_dashboard; print('安装成功')"
📋 安装步骤说明
  1. 访问 GitHub 仓库获取工作流文件
  2. 在对应平台（Dify / Flowise / Make 等）中找到「导入工作流」功能
  3. 上传工作流文件
  4. 按照提示配置必要的环境变量和 API Key
  5. 运行测试确认流程正常后投入使用
以下用法示例由 AI Skill Hub 整理，涵盖最常见的使用场景。
常用命令 / 代码示例
# 命令行使用
claude-agents-dashboard --help

# 基本用法
claude-agents-dashboard input_file -o output_file

# Python 代码中调用
import claude_agents_dashboard

# 示例
result = claude_agents_dashboard.process("input")
print(result)
以下配置示例基于典型使用场景生成，具体参数请参照官方文档调整。
配置示例
# claude-agents-dashboard 配置文件示例（config.yml）
app:
  name: "claude-agents-dashboard"
  debug: false
  log_level: "INFO"

# 运行时指定配置文件
claude-agents-dashboard --config config.yml

# 或通过环境变量配置
export CLAUDE_AGENTS_DASHBOARD_API_KEY="your-key"
export CLAUDE_AGENTS_DASHBOARD_OUTPUT_DIR="./output"
📑 README 深度解析 真实文档 完整度 58/100 含工作流图 查看 GitHub 原文 →
以下内容由系统直接从 GitHub README 解析整理，保留代码块、表格与列表结构。

简介

<a href="https://claude.ai"><img src="made-with-claude.png" height="32" alt="Made with Claude"></a>

<img src="screenshot.jpg" height="420" alt="Screenshot">

Features

  • Kanban board with drag-and-drop (smooth card spacing), create/edit/delete items
  • Save & Start — create an item and immediately launch an agent in one click
  • Start Copy — start a copy of a Todo item while keeping the original, useful for running task variations
  • Agent orchestration via Claude Agent SDK — multiple agents can run simultaneously
  • Git worktrees — each agent works in isolation, branched off main
  • Live work log — streaming agent output via WebSocket (messages, thinking, tool use)
  • Review & merge — tabbed dialog with description, diff viewer, and work log; approve or request changes
  • Clarification flow — agents can ask the user questions mid-task via custom MCP tool; the optional context field on ask_user is rendered as a panel above the prompt so the user sees relevant background before answering
  • Todo creation — agents can create new todo items while working, breaking down complex tasks into smaller actionable items; supports requires parameter to declare dependencies between items and auto_start to automatically launch agents when dependencies are resolved
  • Custom commit messages — agents set meaningful commit messages via MCP tool, used when merging
  • Board introspection — agents can view the current board state (all items by column) via the view_board MCP tool to understand project context
  • Tool access requests — agents can request permission to use optional built-in tools (WebSearch, WebFetch) at runtime via the request_tool_access MCP tool with user approval prompt
  • Done column day grouping — completed items grouped by day (Today, Yesterday, etc.) with collapsible sections, compact title lists, and bulk archive per day group
  • Stats dashboard — real-time header bar showing total cost, token usage, active agents, and items completed today; auto-refreshes every 10 seconds and on WebSocket events
  • Cost & token tracking — agent completion logs USD cost and token consumption (input/output/total) per task, persisted to a dedicated token_usage table
  • System notifications — bell icon in header with badge counter; surfaces MCP server connection failures, agent errors, and warnings; dismiss individually or clear all
  • MCP status monitoring — automatically checks MCP server status after agent session connect; failed/disconnected/needs-auth servers create system notifications
  • Retry with session resume — retry resumes the agent's previous session via session_id, preserving conversation context; falls back to fresh start if no session available
  • Cancel & cancel review — cancel a running agent or discard review changes, clean up worktree/branch
  • Annotation canvas — drop images, scale/move them, draw arrows, circles, rectangles, and text; saved as PNG attachments
  • Attachments — attach annotated screenshots and reference images to items
  • Per-item model selection — choose between Claude Opus 4.8 (default), Opus 4.7/4.6/4.5, Claude Sonnet 4.6, and Claude Haiku 4.5 per item (falls back to global config)
  • Auto-approve modes — per-item setting to skip the manual review gate: OFF (lands in Review for a human), REVIEW (spawns a read-only review agent that auto-merges on approval or sends comments back to the original agent, capped at 3 round-trips), or DIRECT (auto-merge as soon as the agent finishes)
  • Knowledge graph (graphify) — the dashboard owns a navigable AST + optional semantic graph of the target project (graphify-out/); when enabled in Settings ▸ Graphify, agents get a read-only graph_query MCP tool to orient before editing, and the graph auto-refreshes after a merge (free AST build); managed via /api/graphify/* with live build-progress over WebSocket
  • Skills library — a dashboard-managed library of Agent Skills: browse Anthropic's public skills repo, or install any skill from a GitHub repo/path/URL via Settings ▸ Skills. Installed skills are stored in a gitignored skill-library/ (each wrapped as a one-skill plugin) and enabled per project; enabled skills are delivered to agents through the SDK plugins= option so they load regardless of git/worktree/setting_sources. Managed via /api/skills/*
  • Per-task Chrome integration — items can opt into a Chrome browser session for the agent (use_chrome), enabling browser-driven work
  • Transient API-error handling — agent completions classify and persist api_error_status on token_usage, surfacing transient Claude API failures (e.g. overload) distinctly from real agent errors
  • Multi-repo workspaces — point the dashboard at a folder containing sibling git repos and each item picks one of them via a repo field; worktrees are created inside the chosen subrepo and agents get read-only access to the other sibling repos for cross-repo context
  • WIP limit — configurable cap on concurrent running agents; items started beyond the limit are queued and auto-started when a slot opens
  • Agent config — set system prompt, model, project context, MCP servers, and plugins
  • MCP support — connect external tools and data sources via Model Context Protocol; includes an example stdio server (examples/mini-mcp/) for reference
  • Plugin support — load local Claude Code plugins via directory paths
  • Merge conflict auto-resolution — on merge conflict, captures the agent's diff, resets the worktree to the latest base branch, and restarts the agent with the previous diff as context for automated recovery
  • Item cleanup — deleting an item stops running agents, removes worktrees and branches, and cleans up attachment files
  • WebSocket reconnection — automatic reconnection with exponential backoff, visibility-aware, manual reconnect via status indicator
  • WebSocket rate limiting — per-IP connection limits (5 concurrent, 10 per 60s window) prevent resource exhaustion
  • Stats caching — server-side stats caching with 30s TTL, invalidated on mutations for fresh data
  • Git operation timeouts — configurable timeouts for git operations (5min), merges (10min), and HTTP requests (11min)
  • File browser — browse the target project's source code in a full-featured dialog with directory tree, tabbed file viewer, Prism.js syntax highlighting, rendered markdown with mermaid diagrams, inline image previews, secret file hiding, file filter, keyboard navigation, and breadcrumb navigation
  • Allowed commands — configure which shell commands agents can run (e.g., flutter, npm, cargo); agents can request access at runtime via MCP tool with user approval prompt
  • Bash YOLO mode — optional mode that grants agents unrestricted bash access (configurable per project via agent config)
  • Base branch tracking — worktrees record which branch they were created from for reliable merge targeting
  • Base commit pinning — worktrees record the exact commit SHA at creation time, ensuring diffs remain stable even when the base branch moves forward (e.g., after merging other items)
  • Merge commit tracking — stores the merge commit SHA when items are approved, enabling traceability from board items to git history
  • Dirty repo detection — blocks merge if your working tree has uncommitted changes overlapping with the agent's files; moves the item to the Questions column with guidance to commit or stash first
  • Epic grouping — organize items into epics with a collapsible progress panel above the board, colored badges on cards, Todo column grouping by epic, board filtering by epic, inline epic creation in the item dialog, and agent integration via MCP tools; 8 preset colors with light/dark theme variants
  • Auto-start pipelines — items with auto_start enabled automatically launch an agent when all their dependency items are completed, enabling pipeline-style workflows
  • Search — spotlight-style search dialog (Cmd/Ctrl+K) to find items across all columns and search work log entries
  • Archive cleanup — archiving items automatically cleans up their worktree and session resources
  • Shortcut creation — agents can add quick-launch bash command shortcuts to the board via the create_shortcut MCP tool (e.g., test runners, build commands)
  • Shortcuts bar — quick-launch bash commands from a bar at the bottom of the board; commands run as subprocesses with streaming output, stop (preserves output log), reset, auto-reset mode, and cleanup
  • Worktree file browser — browse an agent's worktree files during review via a tree view within the review dialog
  • Retry merge — re-attempt a failed merge without restarting the agent
  • File change detection — when an agent completes, the system detects whether any files were changed; review cards show "Done" (no changes) or "Approve & Merge" (has changes) accordingly
  • Standalone item detail page — each item has a shareable URL; Done detail dialog includes a copy-link button for sharing
  • Animated flame background — optional animated flame effect behind board columns with activity-driven intensity; configurable via agent config (flame_enabled setting)
  • Ollama provider (experimental) — run agents against local Ollama models via Claude Code's env override mechanism; dynamic model discovery, connection status indicator, provider badges on cards; enable with --experimental flag
  • Light/dark mode — respects system preference with manual toggle

Requirements

  • Python 3.12+ (tested on macOS, Linux, and Windows with WSL)
  • Git (any modern version)
  • Claude Code - must be installed and logged in (claude CLI). The dashboard uses the Claude Agent SDK which authenticates through your Claude Code session — no API key needed
  • Internet connection - for Claude API calls

Quick start

From your project repository:

path/to/claude-agents-dashboard/run.sh

Or pass the project path explicitly:

path/to/claude-agents-dashboard/run.sh /path/to/your/project

The server starts at http://127.0.0.1:8000 (auto-increments ports 8000-8019 if busy). Open the dashboard in your browser — on macOS:

open http://127.0.0.1:8000

Your project must be a git repository, or a parent folder containing one or more sibling git repos (multi-repo workspace mode). Requires Python 3.12+.

Example use cases

  • Bug fixes: Create a "Fix login error" item, let an agent analyze logs and implement a solution
  • Feature development: "Add dark mode toggle" → agent updates CSS, templates, and JavaScript
  • Code refactoring: "Extract payment logic to service" → agent reorganizes code while preserving functionality
  • Documentation: "Update API docs" → agent reviews code and updates documentation files
  • Testing: "Add unit tests for user service" → agent analyzes code and writes comprehensive tests
  • Task breakdown: Agents can create follow-up todos like "Add integration tests" or "Update documentation" as they discover related work

API Reference

REST Endpoints

MethodPathDescription
GET/Board page (HTML)
GET/api/itemsList all items
POST/api/itemsCreate item
PATCH/api/items/{id}Update item
DELETE/api/items/{id}Delete item (full cleanup)
POST/api/items/{id}/moveDrag-drop reposition
POST/api/items/{id}/startStart agent
POST/api/items/{id}/cancelCancel agent
POST/api/items/{id}/retryRetry failed agent
POST/api/items/{id}/approveApprove & merge
POST/api/items/{id}/request-changesSend feedback to agent
POST/api/items/{id}/pausePause running agent
POST/api/items/{id}/resumeResume paused agent
POST/api/items/{id}/cancel-reviewDiscard review changes
POST/api/items/{id}/retry-mergeRetry a failed merge
POST/api/items/{id}/start-copyStart a copy of a todo item
POST/api/items/{id}/approve-commandApprove/deny agent command request
GET/api/items/{id}/dependenciesGet item dependencies
PUT/api/items/{id}/dependenciesSet item dependencies
GET/api/items/{id}/is-blockedCheck if item is blocked
GET/api/items/blocked-statusBlocked status for all items
POST/api/items/archive-by-dateBulk archive items by date
POST/api/items/archive-by-epicBulk archive done items by epic
POST/api/items/delete-by-dateBulk delete items by date
POST/api/items/delete-by-epicBulk delete items by epic
GET/api/items/{id}/worktree/treeBrowse worktree directory tree
GET/api/items/{id}/worktree/contentRead file from worktree
GET/api/items/{id}/logWork log entries
GET/api/items/{id}/diffDiff + changed files
GET/api/items/{id}/files/{path}File content at branch
GET/api/items/{id}/clarificationPending clarification
POST/api/items/{id}/clarifySubmit clarification response
GET/POST/api/items/{id}/attachmentsList/upload attachments
DELETE/api/attachments/{id}Delete attachment
GET/api/assets/{filename}Serve uploaded files
GET/PUT/api/configAgent configuration
GET/api/notificationsList system notifications
DELETE/api/notifications/{id}Dismiss a notification
DELETE/api/notificationsClear all notifications
GET/api/statsUsage & activity stats
GET/api/epicsList all epics with progress stats
POST/api/epicsCreate epic
PUT/api/epics/{id}Update epic
DELETE/api/epics/{id}Delete epic (nullifies items' epic_id)
GET/api/config/available-toolsList available optional tools
GET/api/search/worklogSearch work log entries
GET/api/epics/colorsAvailable epic colors
GET/api/shortcutsList shortcuts
POST/api/shortcutsCreate shortcut
DELETE/api/shortcuts/{id}Delete shortcut
POST/api/shortcuts/{id}/runRun shortcut command
POST/api/shortcuts/{id}/stopStop running shortcut (preserves output)
GET/api/shortcuts/{id}/outputGet shortcut output
POST/api/shortcuts/{id}/resetReset shortcut
GET/api/websocket/statsWebSocket connection stats
GET/api/ollama/modelsDiscover local Ollama models (experimental; ?force=true busts cache)
GET/api/graphify/statusKnowledge-graph status: installed/latest version, build-in-progress, graph stats
POST/api/graphify/buildBuild the graph (semantic flag for the LLM layer)
POST/api/graphify/installUpgrade the graphify package in the dashboard venv
GET/api/graphify/queryQuery the graph (q=...)
GET/api/skillsList installed library skills with per-project enabled flags
GET/api/skills/browseList installable skills from a public source (cached)
POST/api/skills/discoverFind every skill (folder with a SKILL.md) in a repo/path
POST/api/skills/installInstall a skill from a GitHub repo/path/URL spec
POST/api/skills/{name}/enabledEnable/disable an installed skill for this project
DELETE/api/skills/{name}Remove a skill from the library
GET/api/files/treeDirectory tree (lazy, depth-limited)
GET/api/files/contentFile content (text, image, binary)
WebSocket/wsReal-time event stream

Troubleshooting

🎯 aiskill88 AI 点评 A 级 2026-06-06

高质量的开源AI工作流项目，值得关注

📚 实用指南（长尾问题）
适合谁
  • 需要 claude-agents-dashboard 解决具体问题的开发者与运营人员
最佳实践
  • 先在测试环境跑通最小用例，再接入生产数据
常见错误
  • API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
  • Python 依赖冲突：建议用 venv / uv 隔离环境
部署方案
  • 云端托管：可放在 Vercel / Railway / Fly.io 等 PaaS 平台
相关搜索
claude-agents-dashboard 中文教程claude-agents-dashboard 安装报错怎么办claude-agents-dashboard 与同类工具对比claude-agents-dashboard 最佳实践claude-agents-dashboard 适合谁用

⚡ 核心功能

👥 适合谁
  • 需要 claude-agents-dashboard 解决具体问题的开发者与运营人员
⭐ 最佳实践
  • 先在测试环境跑通最小用例，再接入生产数据
⚠️ 常见错误
  • API key 直接提交到 git 仓库（请用 .env 并加入 .gitignore）
  • Python 依赖冲突：建议用 venv / uv 隔离环境

👥 适合人群

自动化工程师和运维人员项目经理和业务分析师希望减少重复性工作的专业人士数字化转型团队

🎯 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

✅ 优点
  • +MIT 协议，可免费商用
  • +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
⚠️ 不足
  • −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
⚠️ 使用须知

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。

📄 License 说明

✅ MIT 协议 — 最宽松的开源协议之一，可自由商用、修改、分发，仅需保留版权声明。

🔗 相关工具推荐

browser-use Agent工作流
Agent工作流
LLM资源合集（精选）
精选100+可直接运行的AI Agent和RAG应用集合。包含完整工作流示例、智能代理框架和检索增强生成系统。适合AI开
Graphiti知识图谱引擎
Agent工作流
LangChain AI开发框架
Agent工作流
📚 相关教程推荐
Agent 工作流入门
帮助中心 · AI Skill Hub
Agent 工作流
帮助中心 · AI Skill Hub
AI 写作提效工作流：从 Prompt 设计到批量内容生产
帮助中心 · AI Skill Hub
n8n vs Make vs Zapier 2025：哪款工作流自动化工具最适合 AI 时代
帮助中心 · AI Skill Hub
📰 相关 AI 新闻
Anthropic Claude 最新进展
AI 资讯 · 知识关联
AI Agent 自主化能力最新进展
AI 资讯 · 知识关联
AI Agent 自主化能力最新进展
AI 资讯 · 知识关联
AI 前沿资讯：What are you running on 16Gb V…
AI 资讯 · 知识关联
🍿 AI 圈相关吃瓜
配了5个 MCP 工具，Claude 一个都没用
AI 圈观察
Filesystem MCP 帮 Claude 找文件，找了整个 node_modules
AI 圈观察
我用了 Karpathy 的提示词，Claude 终于不瞎编了
AI 圈观察
🗺️ 相关解决方案
mcp
mcp-server-setup
agent
ai-agent-workflow
workflow
ai-workflow-templates
cli
cli-productivity
🧩 你可能还需要
基于当前 Skill 的能力图谱，自动补全的工具组合
技能寻求者
MCP · Agent · 工作流
DeepCode Agent工作流
MCP · Agent · 工作流
total-agent-memory MCP工具
为Claude Code和Codex CLI提供持久化记忆功能的开源MCP工具。自动提取知识图谱，支持多轮对话上下文保留，适合需要长期记忆和
natively-cluely-ai-assistant — Claude Skill 中文使用文档
免费开源的AI面试助手，实时转录，隐蔽模式，局部RAG，BYOK。无订阅，防止数据泄露。
kb-arena开源AI工作流
kb-arena是开源的AI工作流，用于Benchmark 9 retrieval architectures（向量、上下文、QnA、知识图
CrewAI 多代理协作平台
MCP · Agent · 工作流

❓ 常见问题 FAQ

claude-agents-dashboard 是一款Python开发的AI辅助工具。开源AI工作流：A standalone scrum board that orchestrates Claude agents working on your project。⭐6 · Python 主要应用场景包括：项目管理和协调。
💡 AI Skill Hub 点评

经综合评估，Claude 代理仪表盘 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。

⬇️ 获取与下载
⬇ 下载源码 ZIP

✅ MIT 协议 · 可免费商用 · 直接从 aiskill88 服务器下载，无需跳转 GitHub

📚 深入学习 Claude 代理仪表盘
查看分步骤安装教程和完整使用指南，快速上手这款工具
⚙️ 安装教程 📚 使用教程
🌐 原始信息
原始名称 claude-agents-dashboard
原始描述 开源AI工作流：A standalone scrum board that orchestrates Claude agents working on your project。⭐6 · Python
Topics AI工作流Scrum
GitHub https://github.com/epatel/claude-agents-dashboard
License MIT
语言 Python
🔗 原始来源
🐙 GitHub 仓库  https://github.com/epatel/claude-agents-dashboard

收录时间：2026-06-06 · 更新时间：2026-06-08 · License：MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。