#1010th of 15 in Agent frameworks and SDKs
agentscope
Python agent framework with a multi-tenant FastAPI agent service
- Stars
- 33.2k
- License
- Apache-2.0
- Last commit
- Oct 2026
- Last release
- Oct 2026
- Language
- Python
19,073 on 24 Mar 202633,238 on 13 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
AgentScope 2.0 is a Python 3.11+ SDK for composing ReAct agents from a toolkit (Python tools, MCP servers, skills, shell and file tools), model adapters, context compaction, middleware and permission checks. It also ships an agent service: a FastAPI backend with a sample web UI, multi-tenant sessions, agent teams, Feishu and Discord channels, RAG, scheduling and SQL/NoSQL persistence.
Who it is for: Python developers building agent apps and multi-user agent backends
Strengths
- Wide model support: OpenAI, Anthropic, Gemini, DashScope, DeepSeek, Moonshot, Volcengine, xAI, Ollama, MiniMax
- Sandbox backends include Docker, Bubblewrap, E2B, Daytona, K8s and Apple Container
- Built-in permission system with human confirmation and composable middleware
- Agent service adds multi-tenancy, scheduling, agent teams and A2A support beyond the SDK
Weaknesses
- Realtime voice and SOP features are marked Experimental
- Web UI needs a separate pnpm/Node setup from the Python backend
- README gives no hardware requirements and no Docker or compose files
- 2.0 is a new major version; README does not cover migration from 1.x
What it needs
- no GPU
- Needs Python 3.11+, pnpm (web UI), LLM provider API key
- Models: OpenAI, Anthropic, Gemini, DashScope, DeepSeek
Also in Agent frameworks and SDKs
See all 15| Rank | Project | Score |
|---|---|---|
| 1 | langchainPython framework for building LLM agents and apps from interoperable components | 88 out of 100 |
| 2 | browser-usePython library that lets an LLM agent control a web browser | 88 out of 100 |
| 3 | crewaiPython framework for multi-agent workflows with role-based Crews and event-driven Flows | 80 out of 100 |
| #4 | openai-agents-pythonPython SDK for multi-agent workflows with handoffs, guardrails and tracing | 77 out of 100 |
| #5 | agnoPython SDK and runtime for building and serving agents behind a REST API | 76 out of 100 |