AgentS is an open-source, self-hosted platform designed to manage your AI Agents as true digital assets. It treats an AI's Soul (Persona), Skills (Toolsets), and Memory (Context) as portable, localized assets that you own and control.
- 🛡️ Asset Ownership: All agent data is stored as local Markdown files. Your AI assets stay with you, independent of any specific UI or platform.
- 🧠 Soul & Skill Architecture: Manage complex agent configurations with a structured approach.
- Soul: Core identity and system prompts.
- Skills: Version-controlled toolsets and capabilities.
- Memory: Deep, context-aware knowledge fragments.
- 🚀 One-Click Portability: Effortlessly transfer assets between ChatGPT, Claude, Gemini, or Cursor using our standardized transfer protocol.
- 📚 Prompt Management: A high-performance personal prompt library with full-text search and tagging.
- 🌍 Internationalization: Built-in support for English and Chinese.
- 💎 Premium UI: A "Cyber-Refined" interface built with Next.js 15, Tailwind v4, and Shadcn UI.
AgentS follows a "Data-Centric" architecture where the filesystem is the source of truth, and a lightweight database provides metadata indexing.
- Frontend: Next.js 15 (App Router), Tailwind CSS v4, Milkdown (Markdown Editor).
- Backend: FastAPI (Python), SQLite (Metadata Indexing), Pydantic.
- Storage: Flat Markdown files in
data/for maximum transparency and Git-friendliness.
docker-compose up -d- Open: http://localhost:3000
- API Docs: http://localhost:8000/docs
cd backend
pip install -r requirements.txt
python -m app.maincd frontend
npm install
npm run devYour data is organized transparently:
/data/agents/: Asset folders organized by Agent ID./data/prompts/: Global prompt templates.agents.db: Metadata index for lightning-fast search.
We welcome all contributions! Please see our Contributing Guidelines and Code of Conduct.
This project is licensed under the MIT License.