Preface: Why an AI Gateway Now

Agentic AI is exploding: multiple LLMs, tools, plug‑ins, and data sources. Without a central control plane, teams face key sprawl, inconsistent security, no audit trail, and repeated integrations. An MCP Gateway fixes this by becoming a single entry point where agents discover and use tools, while the platform team enforces:

• Security: RBAC, OAuth/JWT, optional mTLS
• Governance: PII/Secrets filters, schema guards, rate limits
• Observability: JSON logs and OpenTelemetry traces

This site takes you from first run to production‑grade, finishing with a capstone: a CrewAI agent securely calling a Langflow tool through the MCP Gateway, with guardrails and logs that prove it.

What you’ll build and learn

• Stand up an MCP Gateway locally (and optionally with Docker Compose)
• Register servers and federate tools into a single catalog
• Wrap Langflow as a gateway‑managed tool via a tiny adapter
• Write a CrewAI agent that uses the gateway (not raw services)
• Apply guardrails (rate limiters, secret/PII filters, schema guards)
• Enforce role‑based access with RBAC and optional On‑Behalf‑Of (OBO)
• Capture logs & traces for auditability and cost reporting