AgentScope Runtime
https://github.com/agentscope-ai/agentscope-runtime- Category
- AI Agents
- Rank
- No. 775Tools index
- Pricing
- Open Source
- Type
- TOOL
- Builder
- agentscope-ai
- GitHub
- 862 stars
- Latest release
- v1.1.6.post2
- Date
About
Production runtime for agent apps with secure tool sandboxing, Agent-as-a-Service APIs, and full-stack observability.
What it does
AgentScope Runtime wraps Python agents in streaming web services and manages their operating environment. It handles startup, requests, shutdown, deployment state, memory, conversation history, tracing, and isolated execution for browser, filesystem, shell, GUI, and mobile tools.
Why it's ranked here
The scope is unusually complete, spanning agent serving, lifecycle management, deployment, protocols, state services, tracing, and several sandbox types. However, the repository is entering read-only archival status because its capabilities moved into AgentScope 2.0. It is now stronger as a reference or migration source than as a new foundation.
What's good
One package connects local development to Kubernetes and serverless deployment. Services can stream results through server-sent events and expose A2A, Response API, and OpenAI-compatible access. Both synchronous and asynchronous sandboxes cover browser, filesystem, GUI, mobile, training, and cloud use cases. FastAPI integration leaves room for normal web application extensions.
Tradeoffs
The maintainers recommend migration to AgentScope 2.0 and say this repository will become read-only, which sharply limits its appeal for new systems. Framework support is also uneven: AgentScope, Microsoft Agent Framework, and Agno cover messages and tools, while LangGraph lacks completed tool support and AutoGen lacks completed message support.
How to use it well
Use it to understand or maintain an existing Python agent service that needs streaming endpoints, lifecycle hooks, deployment controls, and isolated tools. It also offers a concrete reference architecture for moving agent workloads from local runs toward clustered or serverless environments. For new projects, follow the maintainers toward AgentScope 2.0. It does not replace the agent framework itself.
Technical notes+
pyproject.toml defines a Python 3.10+ setuptools package under src, with FastAPI, Uvicorn, MCP, Docker, Redis, Kubernetes, Celery, A2A, OpenAI, and AgentScope among its core dependencies. It registers six console scripts, including agentscope, runtime-sandbox-mcp, and runtime-sandbox-server. src/agentscope_runtime/cli/cli.py assembles chat, run, web, deploy, list, status, stop, invoke, and sandbox commands. src/agentscope_runtime/engine/__init__.py exports AgentApp and Runner while lazily loading multiple deployment managers. src/agentscope_runtime/sandbox/__init__.py eagerly imports synchronous and asynchronous sandbox classes so registry side effects occur at import time. src/agentscope_runtime/cli/state/__init__.py preserves backward compatibility by re-exporting deployment state types from their newer engine location.
Observed
- License
- Apache 2.0
- Primary language
- Python
- Python requirement
- Python 3.10 or higher
- Packaging
- Setuptools package installable from PyPI with pip or uv; editable source installation is documented
- Interfaces
- Python library, unified CLI, MCP servers, streaming SSE APIs, A2A, Response API, and OpenAI-compatible access
- Deployment targets
- Local, Kubernetes, Knative, Kruise, serverless, ModelStudio, AgentRun, and Function Compute managers are exposed
- Repository status
- Maintainers state that the repository will remain read-only and be archived, with capabilities integrated into AgentScope 2.0
Read from README.md, setup.py, pyproject.toml, src/agentscope_runtime/__init__.py, src/agentscope_runtime/cli/cli.py, src/agentscope_runtime/cli/__init__.py, src/agentscope_runtime/tools/__init__.py, src/agentscope_runtime/engine/__init__.py, src/agentscope_runtime/sandbox/__init__.py, src/agentscope_runtime/cli/state/__init__.py, src/agentscope_runtime/cli/utils/__init__.py, src/agentscope_runtime/engine/app/__init__.py.
What it can do
Execute agent applications in production environment
Agent application code → Running agent application instance
Provide secure tool sandboxing for agent operations
Agent tool requests and code execution → Isolated execution environment with security boundaries
Expose Agent-as-a-Service APIs
API requests to agent endpoints → Agent responses and results via REST/API interface
Monitor agent application performance and behavior
Running agent applications → Performance metrics, logs, and observability data
Track agent execution flows and debugging information
Agent runtime operations → Execution traces, error logs, and debugging insights
Manage agent application deployments
Agent application packages and configuration → Deployed and configured agent services
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.