agentscope-ai/agentscope-runtime
A runtime framework for deploying and orchestrating AI agents at scale with sandboxing, observability, and framework compatibility.

AgentScope Runtime provides a production-ready environment for building and running AI agent applications. It offers secure tool sandboxing to isolate agent executions, Agent-as-a-Service APIs for scalable deployment, and full-stack observability for monitoring agent behavior. The framework is designed to be compatible with multiple agent frameworks including LangGraph and Agno, and supports standard protocols like MCP (Model Context Protocol) and A2A (Agent-to-Agent). Deployment options include Docker and Kubernetes, with serverless capabilities for flexible scaling.