Level 3 · Advanced Scale & Security¶
Goal: go from "one server works" to "several servers work together reliably, safely, and observably" — the engineering that separates a demo from a system you'd trust in production.
Modules¶
- Multi-Server Clients: Connecting to Several Servers at Once
- Tool Selection & Namespacing at Scale
- Error Handling & Resilience Patterns
- Streaming Responses & Progress Notifications
- Roots: Scoping a Server to a Workspace
- Security: Prompt Injection & Tool Poisoning Risks
- Scopes, Permissions & Least Privilege
- Caching & Rate Limiting Tool Calls
- Observability: Logging & Tracing MCP Calls
- Project — Orchestrating 3 MCP Servers for a Research Agent
By the end of this level you'll be able to wire up and secure multi-server agent systems, reason about MCP-specific security risks like prompt injection and tool poisoning, and keep a multi-server setup observable and resilient under real failures.