Lightweight, scalable, reliable agentic AI framework, built for imaginations.
Website • Documentations • Market • Telegram Community • YouTube
-
🌌 Omniscient & Omnipotent. Supports OpenAI-standard Skills, featuring Generative Skills, Workflows, Computer control, and more.
-
⚡ Lightweight yet reliable. Completely developed with Go and HTML for maxmized speed, while supporting multi-threading for real-time execution.
-
💻 Bash for LLM. Utilizing pseudo-terminals and natural-language-like programming, empowering your Agent to use Skills and APIs via LLMs out-of-the-box.
-
🛡️ Privacy by Design. Leveraging a revolutionary Edge + Cloud hybrid model architecture to provide privacy-first desensitization and semantic obfuscation capabilities.
-
🗜️ Multi-level Context Compression. Utilizing Generative RAG, Preferences and highly configurable parameters to grant your Agent a long-term, stable personality and memory.
-
🔐 Granular Permission Control. Let your Agent know what it can do, and respect what it cannot. Strictly manage the Agent's capability boundaries via the Sources and Outputs panels.
-
☁️ Privacy-First Cloud Features. Access and download Skills, Workflows, and Agent presets for free via the companion Matex Market; allow authorized third-party apps to access data securely through the Sync mechanism.
-
🤝 Multi-Model Collaboration. Supports multi-model cooperation, multi-modal synergy, and model orchestration, allowing your Agent to combine the best of all worlds.
-
🔌 Choose Your Favorite Provider. Fully compatible with OpenAI-standard APIs, as well as mainstream model providers like Ollama, LM Studio, Phala, and more.
-
📱 Choose Your Favorite Client. Use our open-source Agent clients, or interact with your Agent directly via Web GUI or Telegram Bot.
-
🛠️ Development Made Simple. Comes with comprehensive API and SDK documentation for you to effortlessly add brand new capabilities and sensory data sources to your Agent.
We believe in absolute simplicity. Deploying Zuver takes just a few steps:
Head directly to the Releases page to download the pre-compiled build corresponding to your system architecture. Run it directly on a machine with a Go environment.
If you want to experience the bleeding-edge features or participate in development, simply clone the repository and spin it up:
# Clone the repository
git clone https://github.com/synasties/zuver.git
cd zuver
# Run it (as simple and elegant as that)
go run main.go
Want to explore advanced configurations, write custom Skills, or dive deep into the API structure? 👉 Click here to visit GitHub Documentations
This project is licensed under the AI-Involved Non-Commercial & Responsible Public License (AI-NCRL) v1.1. We believe in the sustainability of open source. To protect our ability to innovate while giving back to the community, every version of this software will automatically convert to the Apache License 2.0 exactly six (6) months after its release.
- Hobbyists & Researchers: Free for non-commercial use! 🚀
- Commercial Use (Current Version): Requires a commercial license. 💼
- Commercial Use (Versions > 6 months old): Already Apache 2.0, enjoy! 🎉
(Please see the
LICENSEfile for full details and restrictions).
If you feels it's not wise to commit, no worry. Just leave us with the idea, and we will get it into reality really soon!
The Zuver name, logos (including the visual mark featured above), and related brand assets are the exclusive trademarks of the Project Owner. While the source code is available under the specified licenses, you may NOT use our branding or logos in any derived, forked, or hosted projects without explicit permission.
Disclaimer: All trademarks, logos and brand names mentioned in this project are the property of their respective owners. All company, product and service names used in this repository are for identification purposes only. Use of these names, trademarks and brands does not imply endorsement, sponsorship, or affiliation.