[General] Why should I use AgentScope Runtime instead of plain LangGraph + uvicorn?

[flyflypeng]

I have already built several agent applications with LangGraph SDK. LangGraph gives me:

• Built-in execution state tracking
• Full conversation history
• Long-term memory (via the LangMem extension)

When I need to go live I simply wrap the app with a thin uvicorn layer, build a container image, and deploy it to my Kubernetes cluster.

Given that background, I’m struggling to see what extra value the AgentScope Runtime would add for me.

Could you:

1. Explain the concrete benefits I would gain by switching to (or adding) AgentScope Runtime instead of sticking with my current LangGraph + uvicorn + K8s workflow?
2. Provide a minimal end-to-end example that shows how to run an existing LangGraph-based agent under AgentScope Runtime?

Having a runnable sample would help me evaluate whether the migration is worth it.

[rayrayraykk]

Hi @flyflypeng, thanks for your interest in AgentScope‑Runtime.

If you already work with LangGraph + uvicorn + K8s, the extra value AgentScope‑Runtime can provide is:

1. Protocol support
   • Directly support OpenAI SDK and Google A2A — allowing your agent to become an API that works with standard clients and other agents without extra integration work.
2. Built‑in tools & sandbox
   • Secure, ready‑to‑use environments (code, browser, filesystem, GUI) and utilities like Search, RAG, AIGC, Payment to help you build and run agents more easily.
3. Works within your existing workflow
   • Import Runtime as a library, keep your current app and deployment process, just wrap it to gain sandbox and tool support.
4. One‑click Alibaba Cloud deployment
   • Skip manual Docker image builds, get a Web UI automatically after you push.
5. Native runtime of AgentScope
   • AgentScope‑Runtime is the native runtime framework of [AgentScope](https://github.com/agentscope-ai/agentscope), an agent development framework. You can try AgentScope itself to explore more advanced agent development capabilities.

For quick start and sample code:

• Quickstart guide: https://runtime.agentscope.io/en/quickstart.html
• Browser‑use agent full‑stack sample: https://github.com/agentscope-ai/agentscope-samples/tree/main/browser_use/browser_use_fullstack_runtime

[flyflypeng]

Alright, your project offers valuable insights for advancing Agent applications from demos to production deployment. Also, based on the GitHub name I saw, I guess you're a fan of the Ray project—so wouldn't it be interesting to use it as an Agent Runtime?

We’d still like to see more examples of integrating various Agent frameworks into the AgentScope Runtime—it would be far more straightforward.