Model Context Protocol Demo

A real MCP server for weather data, with tools, resources, prompts, and live testing.

This app exposes a Streamable HTTP MCP endpoint at /mcp and a professional dashboard that explains how MCP works.

Test MCP Learn concepts

MCP endpoint

http://localhost:8000/mcp

Dashboard wrappers

/api/mcp/tools/call
/api/mcp/resources/read
/api/mcp/prompts/get

Tools

Executable functions that an AI client can call when it needs to take action or fetch live data.

tools/call

Resources

Read-only context exposed through URIs. Good for files, configs, profiles, docs, and reference data.

resources/read

Prompts

Reusable templates that help clients ask better questions or produce consistent outputs.

prompts/get

Why MCP is required

Without MCP, every AI client needs custom integration code for every backend. MCP gives a common contract for discovery, calling tools, reading context, and reusing prompts.

Problem	MCP solution
Different AI clients integrate differently	One standard protocol
Tools are hard to discover	`tools/list` exposes names, descriptions, and schemas
Context gets copied manually	`resources/read` exposes data on demand
Prompts are scattered	`prompts/list` and `prompts/get` make templates reusable

Observability Architecture

This demo is instrumented with production-grade telemetry ready for the Grafana stack.

Logs

stdout → Grafana Alloy → Loki → Grafana

Metrics

/metrics → Prometheus → Grafana

Traces

OpenTelemetry OTLP → Alloy → Tempo → Grafana

Health

/health | /ready

Live MCP tester

These forms use the FastMCP client in-memory, so you are testing the actual registered MCP tools/resources/prompts.

Tool

Arguments JSON

Resource URI

Prompt

Arguments JSON

Click refresh catalog to list what the MCP server exposes.

Result

Ready.