---

name: service-platform description: "AgentScope FastAPI service, storage, hub, channel, MCP, and deployment workflows." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

service-platform

Use this sub-skill for the AgentScope service layer: create_app, storage, message bus wiring, workspace managers, hub registration, channel integration, and deployment/bootstrap checks.

Read first

• references/platform-overview.md for the create_app contract and a bootstrap sketch.
• references/storage-and-bus.md for storage, message-bus, workspace-manager, channel, hub, and MCP components.
• references/troubleshooting.md for service startup and backend failures.
• scripts/service_smoke.py for a safe local bootstrap check.

Typical triggers

• Start or debug the AgentScope FastAPI service.
• Compare InMemoryMessageBus vs RedisMessageBus.
• Wire RedisStorage or AsyncSQLAlchemyStorage into create_app.
• Add workspace managers, skill hubs, MCP hubs, or channel backends.
• Diagnose index-worker, knowledge-base, or service startup problems.

What belongs here

• agentscope.app.create_app
• SubAgentTemplate
• message bus, storage, workspace manager, knowledge-base manager
• MCP client, channel, and hub integration
• service bootstrap, deployment, and runtime checks

What does not belong here

• agent/tool/permission basics → agent-core
• provider credentials and model classes → provider-connectors
• retrieval or memory workflows → rag-memory
• workspace backend internals → workspace-sandboxes

Use pattern

1. Decide whether the problem is storage, message bus, workspace, or channel related.
2. Read the component reference that matches that layer.
3. Start with the local smoke script before changing a live deployment.
4. Add external backends only after the in-memory or local bootstrap works.
5. Escalate to workspace-sandboxes if the failure is really a sandbox backend issue.

Cross-links

• If the issue is actually a provider, use provider-connectors.
• If the issue is a tool or agent loop problem, use agent-core.
• If the issue is retrieval or memory, use rag-memory.