Demos

This is a working demonstration of governed agentic AI built on Gravitee. The scenario is an RNLI lifeboat station finder — a real agent, real policies, real governance. Everything you see is portable and transferable to any use case.

What this demo covers:

• A real agent query — finding the nearest lifeboat stations — with every hop visible in the Agent Inspector in real time. The observability comes from the gateway, not the agent code
• MCP tool calling through Gravitee — the gateway natively transforms REST APIs into AI-discoverable tools, no agent instrumentation required
• A2A agent delegation — the station finder hands off to a separate Sea Conditions agent for tidal and weather data
• Guard rails in action — a harmful prompt is classified and blocked before it reaches the LLM; a prompt injection attempt is caught by the injection screening policy. Both blocked at the gateway
• Plan-gated access — logging in as an authenticated Gold plan user unlocks visit history, richer sea condition data and live Kafka event streams via SSE. The same agent, different data, controlled entirely at the gateway
• The full governance wrapper — rate limiting, guard rails, plan-gated access and observability. None of it lives in the agent code