Agent Runtime (AGR) is a next-generation infrastructure platform designed around the native execution model of AI agents.
Unlike traditional request–response applications, AI agents are goal-oriented, autonomous, multi-step, and tool-driven. These execution characteristics introduce new infrastructure challenges that conventional cloud runtimes were never designed to solve.
Agent Runtime rethinks how AI agents are deployed, executed, managed, and governed. Built for Agentic RL, Agentic AI applications, and enterprise agent platforms, it provides a comprehensive runtime stack including:
- 🚀 Agent Runtime
- 📦 Secure Sandbox
- 🔧 Tool Infrastructure
- 🌐 Agent Gateway
- 🧠 Memory and Session
- 📊 Observability
- 🔒 Governance & Security
By providing deterministic infrastructure for inherently non-deterministic AI agents, Agent Runtime enables developers to build reliable, scalable, and production-ready agent systems.
Explore the resources below to start building with Agent Runtime.
| Resource | Description |
|---|---|
| 📖 Cookbook | End-to-end examples and tutorials |
| 💻 AGR CLI | Manage agents and sandboxes from your terminal |
| 🌐 Product Website (CN) | Learn about Tencent Cloud Agent Runtime |
| 📚 Documentation (CN) | Product documentation and developer guides |
We welcome feedback, feature requests, and contributions. If you'd like to report an issue or contribute to one of our projects, please open an issue or pull request in the corresponding repository.