能力标签

🔌 MCP 🤖 Agent 🔄 工作流 💻 CLI 🔗 REST API 🧠 Claude

⚙️

Agent工作流

开源AI工作流

基于 TypeScript · 无代码搭建完整 AI 自动化流程

英文名：oss-autopilot

⭐ 9 Stars 🍴 3 Forks 💻 TypeScript 📄 MIT 🏷 AI 8.0分

8.0AI 综合评分

ai-agentclaude-codetypescript

⬇ 下载源码 ZIP ⚙️ 配置说明

✦ AI Skill Hub 推荐

AI Skill Hub 强烈推荐：开源AI工作流是一款优质的Agent工作流。AI 综合评分 8.0 分，在同类工具中表现稳健。如果你正在寻找可靠的Agent工作流解决方案，这是一个值得深入了解的选择。

📚 深度解析

开源AI工作流是一套完整的 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。

开源AI工作流工作流的设计遵循"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 8.0 分，是同类 Agent 工作流中的精选推荐。

📋 工具概览

开源AI工作流是一套完整的 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

⭐ 9

开发语言

TypeScript

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

8.0 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」。

📌 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：npm 全局安装
npm install -g oss-autopilot

# 方式二：npx 直接运行（无需安装）
npx oss-autopilot --help

# 方式三：项目依赖安装
npm install oss-autopilot

# 方式四：从源码运行
git clone https://github.com/costajohnt/oss-autopilot
cd oss-autopilot
npm install
npm start

📋 安装步骤说明

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

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
oss-autopilot --help

# 基本用法
oss-autopilot [options] <input>

# Node.js 代码中使用
const oss_autopilot = require('oss-autopilot');

const result = await oss_autopilot.run(options);
console.log(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

# oss-autopilot 配置说明
# 查看配置选项
oss-autopilot --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 OSS_AUTOPILOT_CONFIG="/path/to/config.yml"

📑 README 深度解析真实文档完整度 52/100 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

An AI-powered workflow engine for managing open source contributions at scale — built as a Claude Code plugin, MCP server, and standalone CLI.

---

Built and used daily by costajohnt — 3rd biggest contributor to Ink (the React CLI framework behind Claude Code, Gemini CLI, and Codex — 32k+ stars) and repeat contributor to Homebrew.

---

Engineering Highlights

┌──────────────────────────────────────────────────┐
│  Claude Code Plugin Layer                        │
│  8 slash commands (/oss, /oss-search, …)         │
│  7 specialized agents, contribution skills       │
├──────────────────────────────────────────────────┤
│                                                  │
│  ┌──────────────┐  ┌──────────────────────────┐  │
│  │ MCP Server   │  │ Interactive Dashboard     │  │
│  │ @oss-auto-   │  │ @oss-autopilot/dashboard │  │
│  │ pilot/mcp    │  │ Preact + Vite             │  │
│  │              │  │ PR management, charts,    │  │
│  │ 32 tools     │  │ actions                   │  │
│  │ 6 resources  │  │                           │  │
│  │ 4 prompts    │  │                           │  │
│  └──────┬───────┘  └────────────┬─────────────┘  │
│         │                       │                │
│  ┌──────┴───────────────────────┴─────────────┐  │
│  │ Core Library — @oss-autopilot/core         │  │
│  │ PR monitoring, issue discovery, state mgmt │  │
│  │ GitHub API, CLI, structured JSON output    │  │
│  └────────────────────────────────────────────┘  │
│                                                  │
└──────────────────────────────────────────────────┘

Monorepo with three npm packages — pnpm workspaces with each package independently publishable to npm. Core library, MCP server, and interactive Preact dashboard with shared types.

Three deployment models — Claude Code plugin with 7 specialized agents, MCP server for Cursor/Claude Desktop/Codex/Windsurf, and a standalone CLI with --json structured output. Same core, different interfaces.

Deterministic core, AI orchestration layer — Critical logic (PR status classification, CI failure analysis, state management) lives in tested TypeScript, not in prompts. The CLI returns structured JSON that agents consume. CI failures are categorized into a deterministic taxonomy — actionable vs. fork limitation vs. auth gate vs. infrastructure — rather than asking an LLM each time. 2,600+ tests validate the core independently of any LLM.

Production-grade GitHub API integration — ETag-based HTTP caching, automatic rate limit backoff with retries, bounded concurrency pools, and paginated fetching. Handles the full complexity of fork-based contribution workflows: correct diff ranges, squash commit counting, and --head flag handling for cross-fork PRs. Designed to run daily without hitting API limits.

Human-in-the-loop guardrails — Nothing is posted to GitHub without explicit approval. AI drafts responses but the contributor always reviews before sending. Pre-commit review gates catch issues before they reach maintainers. Factual claims in draft comments are verified against the actual diff before presenting to the user.

Modular extraction — Issue discovery and vetting grew complex enough to extract into its own npm package (oss-scout). Connected via a bridge pattern that maps state between the two systems, following the same approach used by the broader ecosystem of extraction-and-reconnect patterns.

Fresh-fetch architecture — PRs aren't stored locally. Every run fetches live data from GitHub's Search API and enriches each PR with CI status, review decisions, merge conflict detection, maintainer comment classification, and checklist completion. No stale data, no sync bugs.

Security discipline — State files written with 0o600 permissions, data directory created with 0o700. Concurrent state write protection prevents corruption from parallel runs. Runtime schema validation via Zod on every state file read. XSS prevention tested. Input validation hardened across CLI arguments and API responses.

Automated release pipeline — Conventional commits feed into release-please for automatic versioning and changelogs, with CI/CD publishing to npm on merge. 200+ changelog versions across both packages (core v0.1.0 through v3.x, mcp through v5.x) since the first release in January 2025.

Every feature in the list above was driven by real usage — capacity warnings came from overcommitting, "skip comment when code speaks for itself" came from over-commenting, diminishing returns detection came from spending too long searching. The tool is shaped by the contributions it manages.

---

Install & Usage

Claude Code Plugin (recommended):

/plugin marketplace add costajohnt/oss-autopilot
/plugin install oss-autopilot@oss-autopilot

Restart Claude Code, then run /setup-oss. Done.

<details> <summary>MCP Server (Cursor, Claude Desktop, Codex, Windsurf)</summary>

First initialize your GitHub username (one-time setup):

npx @oss-autopilot/core@latest init <your-github-username>

Then add to your MCP client config:

{
  "mcpServers": {
    "oss-autopilot": {
      "command": "npx",
      "args": ["@oss-autopilot/mcp@latest"]
    }
  }
}

The MCP server exposes 32 tools, 6 resources, and 4 prompts — the full OSS Autopilot feature set.

</details>

<details> <summary>Standalone CLI / npm package</summary>

```bash

Or install globally

npm install -g @oss-autopilot/core

Rebuild

cd <path-from-find-command>/packages/core npm install npm run bundle


**PRs not showing up:**
- Run `/setup-oss` to ensure your GitHub username is configured
- Check that `gh auth status` shows you're authenticated
- The plugin only tracks PRs you authored

</details>

<details>
<summary><strong>Development</strong></summary>

bash git clone https://github.com/costajohnt/oss-autopilot.git cd oss-autopilot pnpm install # Install all workspace dependencies pnpm test # Run all tests across all packages pnpm start -- daily --json # Run CLI via tsx (no bundle needed) pnpm run bundle # Rebuild CLI bundle (esbuild)


**Project structure:**

├── commands/ # 8 plugin slash commands (/oss, /oss-search, /pr-ready, …) ├── agents/ # 7 specialized agents (PR responder, issue scout, etc.) ├── skills/ # Contribution best practices ├── workflows/ # Delegated logic loaded by commands on demand ├── hooks/ # Plugin hooks (session-start) ├── packages/ │ ├── core/ # @oss-autopilot/core — CLI + core library │ │ ├── src/commands/ # CLI subcommands │ │ ├── src/core/ # Domain logic + tests │ │ └── dist/cli.bundle.cjs # Built bundle (auto-generated) │ ├── mcp-server/ # @oss-autopilot/mcp — MCP server │ │ └── src/ # Tools, resources, prompts, server │ └── dashboard/ # @oss-autopilot/dashboard — Interactive UI └── pnpm-workspace.yaml # Workspace definition


Test as a local plugin:

bash claude --plugin-dir ./oss-autopilot ```

</details>

<details> <summary>Enhanced Code Review (optional)</summary>

The plugin includes a built-in pre-commit-reviewer agent that reviews all code changes before committing. For enhanced parallel review, install the pr-review-toolkit plugin (search for it in the Claude Code plugin marketplace) — it adds 5 specialized reviewers that run simultaneously:

Agent	Focus
`code-reviewer`	Bugs, logic errors, security, conventions
`silent-failure-hunter`	Error handling gaps, swallowed errors
`code-simplifier`	Dead code, unnecessary complexity
`pr-test-analyzer`	Test coverage and assertion quality
`comment-analyzer`	Comment accuracy and maintainability
`type-design-analyzer`	TypeScript type design (encapsulation, invariants, enforcement)