@FakeMaidenMaker: 单一 Agent 用来生成图片、做数据分析、剪视频——每一项看起来能干,但想出一份完整的结果,往往得在 ChatGPT、Midjourney、Seedance、Excel 之间来回切五次。 给大家分享一个最近在用的开源项目 VRSEN/O…
摘要
介绍开源多智能体系统 OpenSwarm,8 个专业 agent 协同工作,从单个 prompt 在终端中生成幻灯片、研究报告、数据可视化等完整交付物,无需 GUI。
查看缓存全文
缓存时间: 2026/05/24 00:17
单一 Agent 用来生成图片、做数据分析、剪视频——每一项看起来能干,但想出一份完整的结果,往往得在 ChatGPT、Midjourney、Seedance、Excel 之间来回切五次。
给大家分享一个最近在用的开源项目 VRSEN/OpenSwarm,8 个专业 agent 协同 + 一个 prompt 出完整交付物,全程跑在终端里,不用开任何 GUI。
GitHub:https://github.com/VRSEN/OpenSwarm
- Orchestrator 统一收请求自动分发给对应专家;
- Deep Research 调研竞品顺手带引证;
- Slides Agent 直接生成 HTML deck 并导出 PPTX;
- Data Analyst 跑在隔离的 IPython kernel 里出可视化图表;
- Image 和 Video 接了 Gemini、Sora、Veo 多个模型。
- Virtual Assistant 还通过 Composio 接了 1 万+ 外部服务,Gmail / Slack / GitHub 全打通。
主要功能: Orchestrator 总调度——所有请求路由到对应专业 agent,不自己瞎猜 Deep Research——带引证的网页调研,输出带参考来源的平衡分析 Slides Agent——一个 prompt 生成 HTML slide deck,自动导出 PPTX Data Analyst——在隔离 IPython kernel 里跑统计模型、出可视化图表 Image + Video 多模型——Gemini / http://fal.ai / Sora / Veo 可选,支持视频拼接编辑
部署也很简单,npm install -g @vrsen/openswarm 一行装好,openswarm 起 setup wizard 全自动配好环境。
适合被多工具来回切换烦透的朋友收藏使用。
VRSEN/OpenSwarm
Source: https://github.com/VRSEN/OpenSwarm
🚀 OpenSwarm

The fully open-source multi-agent system that does everything Claude Code can’t.
Create polished slide decks, research reports, data visualizations, documents, images, and videos — all from a single prompt in your terminal. No platform, no UI, no setup hassles.
✨ One prompt → Complete deliverables
🎯 8 specialized agents working together
⚡ Install in 30 seconds, running in 60
🔧 100% customizable and forkable
Built on Agency Swarm — the framework powering real AI swarms.
💼 Investor or looking to integrate AI agents into your SaaS? We’re the team behind OpenSwarm and Agency Swarm, building the future of multi-agent systems. Partner with us →
💡 What Makes This Different?
Instead of one agent trying to do everything poorly, you get specialists coordinated by an orchestrator.
🎯 Real Examples
Paste these into your terminal and watch magic happen:
- “Create a complete investor pitch for OpenSwarm” → Full deck + executive summary + market research
- “Research my top 5 competitors and write 3 SEO-optimized blog posts” → Competitive analysis + keyword research + publish-ready content
- “Analyze this data and create a quarterly report with charts” → Data insights + visualizations + formatted document
- “Generate a product launch video with animations” → Professional video with graphics and transitions
- “Build me a marketing campaign for Q2” → Strategy doc + creative assets + implementation timeline
Connect to 10,000+ external services (Gmail, Slack, GitHub, HubSpot) via Composio for even more power.
🤖 Meet Your AI Team
| Agent | What it does |
|---|---|
| Orchestrator | Routes every user request to the right specialist(s). Never answers directly — pure coordination. |
| Virtual Assistant | Handles everyday tasks: writing, scheduling, messaging, task management. Gains 10,000+ external integrations via Composio (Gmail, Slack, GitHub, HubSpot, and more). |
| Deep Research | Conducts comprehensive, evidence-based web research with citations and balanced analysis. |
| Data Analyst | Analyses structured data, builds charts, runs statistical models — all inside an isolated IPython kernel. |
| Slides Agent | Generates complete, visually polished HTML slide decks, then exports them to PPTX. |
| Docs Agent | Creates formatted Word documents and PDFs from outlines or raw content. |
| Image Generation Agent | Generates and edits images using Gemini 2.5 Flash Image / Gemini 3 Pro Image and fal.ai. |
| Video Generation Agent | Produces videos via Sora (OpenAI), Veo (Google), and Seedance (fal.ai); also edits and combines clips. |
📦 Get Started in 30 Seconds
For most users (recommended):
npm install -g @vrsen/openswarm
openswarm
That’s it! The setup wizard handles everything: authentication, dependencies, and configuration.
Requirements: Node.js 20+ (Python 3.10+ auto-installed)
🔧 Build Your Own Swarm
Fork this repo and create your own specialized AI team in minutes:
git clone https://github.com/VRSEN/openswarm.git
cd openswarm
Then tell Claude Code, Cursor, or Codex:
“Turn this into an SEO optimization swarm”
They’ll automatically customize all agents for your use case.
Popular custom swarms:
- SEO Swarm: Keyword research + competitor analysis + blog writing
- Sales Swarm: Lead research + outreach + proposal generation
- Marketing Swarm: Campaign planning + creative assets + analytics
- Product Swarm: Market research + feature specs + launch materials
⚙️ API Keys & Setup
The setup wizard walks you through everything, but you’ll need at least one of these:
Required (choose one):
OPENAI_API_KEY- For GPT 5.5 and Sora video generationANTHROPIC_API_KEY- For Claude models
Optional superpowers:
COMPOSIO_API_KEY- Unlock 10,000+ integrations (Gmail, Slack, GitHub, etc.)GOOGLE_API_KEY- Gemini image generation + Veo videoFAL_KEY- Advanced video editing and effectsSEARCH_API_KEY- Web search for research agent
Tools gracefully degrade when keys are missing — you’ll get clear instructions on what to add.
🚀 Coming Soon
- Agent Builder Agent - Create custom swarms from a single prompt
- OpenClaw + Claude Code integration - All agents in one place
⭐ Star us on GitHub to stay updated and help us prioritize features!
🏗️ For Developers
Local development:
git clone https://github.com/VRSEN/openswarm.git
cd openswarm
python swarm.py
Docker deployment:
git clone https://github.com/VRSEN/openswarm.git
cd openswarm
cp .env.example .env # Add your API keys
docker-compose up --build
API server:
python server.py # Runs on localhost:8080
📺 Learn More
- Watch the full demo: YouTube video →
- Multi-agent framework: Agency Swarm
- Terminal UI for Agency Swarm: AgentSwarm (OpenCode-based TUI)
- External integrations: Composio
👥 Team
- Artemii Shatokhin — Built the core OpenSwarm agent team: the specialist agents, orchestration layer, shared tools, and runtime integrations. (GitHub)
- Nick Bobrowski — Built the foundation OpenSwarm builds on: Agency Swarm and the AgentSwarm CLI/TUI, an OpenCode-based terminal experience customized for Agency Swarm. (GitHub)
📄 License
MIT — see LICENSE.
Built with ❤️ by the team behind Agency Swarm
相似文章
@justloveabit: 用这个开源工具,我让一群AI替我上班了 事情是这样的,最近一直在折腾各种AI agent。Claude Code开一堆窗口,Codex也在跑,偶尔还要用Cursor。结果呢,乱成一锅粥——哪个agent在干啥,花了多少钱,完全搞不清楚。重…
本文介绍了一款名为Paperclip的开源工具,用于统一管理和调度多个AI Agent。它通过模拟公司组织架构、任务分配与预算控制等功能,解决了多Agent协作时上下文丢失、成本不可控和调度混乱的痛点。
@FakeMaidenMaker: 体量小而可读性高的轻量个人 AI agent 现有 agent 框架把核心 loop 包成黑盒,想改一个细节要扒好几层抽象,读起来都费劲,更别说二次开发了。 我给大家安利一个超轻量开源 AI agent — nanobot。专门为「想读懂…
介绍了开源超轻量AI agent nanobot,其核心loop设计极简可读,易于二次开发,支持Telegram、Discord、飞书等10+聊天平台及多个AI provider,并集成长期记忆与MCP工具扩展。
@QingQ77: 一套能自己部署的 agentic AI + BI 平台:建 agent、跑多 agent swarm、接知识库和数据源、看完整执行轨迹,全部跑在自己机器和自己的 API Key 上。 https://github.com/AgentSwa…
介绍 AgentSwarms,一套可自行部署的 agentic AI 与 BI 平台,支持构建 agent、运行多 agent swarm、接入知识库与数据源,并可查看完整执行轨迹,数据与 API key 均掌握在用户手中。
@seclink: 最近这个开源工具挺火的。 看起来像是 钉钉悟空 、 字节 aily的开源版本。 你可以基于它来实现自己的agent 并且接入到上述的 即时通讯平台之中。 有的哥们基于这个改吧改吧,就能给投资人演示,拿到了不小规模的估值 。 让投资人记忆深…
CowAgent 是一个基于大模型的开源 AI 助理框架,支持自主任务规划、长期记忆、知识库、多模型切换和多渠道接入(微信、飞书、钉钉等),可快速构建和部署个性化 AI agent。
@QingQ77: 本地优先的多智能体协作桌面应用,把 AI 协作做成聊天一样的体验,支持多智能体任务分发、文件审查和人工审批。 https://github.com/lizyoko9/bitdance-agenthub… 一个基于 Next.js + El…
AgentHub 是一个本地优先的多智能体协作桌面应用,将 AI 协作做成聊天体验,支持任务分发、文件审查和人工审批,基于 Next.js 和 Electron 构建。