Week 16 • Day 5Week 16

Day 5:Days 5–7: Month 4 Capstone Project — Autonomous Agentic MCP Server & Infrastructure Copilot

Battle-tested daily lesson note • 100% runnable code, interview Q&A, test suite & interactive quiz

Days 5–7: Month 4 Capstone Project — Autonomous Agentic MCP Server & Infrastructure Copilot

Capstone Location

All production source code, architecture documentation, and automated verification tests for the Month 4 Capstone reside in:

  • capstones/month04_agentic_mcp_server/README.md (open the matching Capstone project in the left sidebar or use Tab 2: Codes below)
  • capstones/month04_agentic_mcp_server/src/agentic_mcp_copilot.py (available in Tab 2: Codes)
  • capstones/month04_agentic_mcp_server/tests/test_capstone_month04.py (available in Tab 4: Test & Troubleshooting)

What You Will Integrate Across Month 4 (Weeks 13–16)

  1. Week 13 (Stateful Graphs, Checkpointing & HITL): SQLite checkpoint persistence and Human-in-the-Loop interrupt/resume gates for mutating production operations.
  2. Week 14 (Model Context Protocol Servers, Semantic Routing & Security): Multi-server JSON-RPC 2.0 MCP Host (observability, kubernetes), Top-K Semantic Tool Router, and path-traversal/secret-redacting sandbox.
  3. Week 15 (Multi-Agent Supervisor, Generator-Critic & Budget Circuit Breakers): Supervisor domain routing, adversarial SecurityAndSRECritic verification, and per-request USD/token/oscillation circuit breakers.
  4. Week 16 (AST Sandbox, Policy-as-Code Rails, OpenTelemetry Tracing & Chaos Fallbacks): 3-Rail Policy-as-Code enforcement, AST-validated metric computation, tamper-evident OpenTelemetry audit spans, and Chaos failover.

In Simple English: Your Month 4 Autonomous Agentic MCP Capstone

In Month 4, you leveled up from stateless chatbots to safe, observable, multi-step AI Agents:

  1. Typed MCP JSON-RPC Tool Discovery + Path-Jail Sandboxing
  2. ReAct / Graph State Machine with Checkpointing + HITL Pause Gates
  3. Policy-as-Code Guardrails + Cost Caps + OpenTelemetry Hierarchical Spans

Interactive Browser Sandbox Included!

Open Tab 2 (Codes) and Tab 4 (Test & Troubleshooting) — we added a complete pure-Python runnable Agentic MCP Copilot Sandbox (agentic_mcp_sandbox.py & test_agentic_mcp_sandbox.py) covering tool-call JSON-RPC dispatch, path-traversal blocking, step budget caps, and structured span tracing right in your browser at $0 server cost!