AgentScope

Use it when

• I need to build a Python LLM agent with a model, prompt, and tools without implementing the execution loop myself.
• My agent conversations exceed the practical context window and need built-in context compaction or tool-result offload.

What it solves

Not the fit when

• A ready-to-use end-user chat application without configuring an agent and model.
• Eliminating the need for credentials or costs from a selected hosted model provider.
• model-provider API costs
• hosted agent-service deployment
• application-specific tool permissions

Install

uv pip install agentscope

Invoke

Create an Agent with a configured model and Toolkit, then iterate `agent.reply_stream(UserMsg("user", "prompt"))` inside `asyncio.run(main())`.

Alternatives

No reviewed alternatives recorded yet.