Ramakrishna-Gedala/mcp-web-knowledge-coach
If you are the rightful owner of mcp-web-knowledge-coach and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.
This repository contains a streaming Model Context Protocol server implemented in Python and a React web client for demos and training on MCP concepts.
Knowledge Coach MCP Demo
This repository contains a streaming Model Context Protocol server implemented in Python and a React web client that interacts with it over the Streamable HTTP transport. It is designed for demos and for training new learners on MCP concepts.
Project layout
.
├── server/ # Python MCP server (FastMCP + FastAPI)
└── web/ # React + Vite chat client
Prerequisites
- Python 3.11+
uv≥ 0.9- Node.js 20+
Setup
1. MCP server
cd server
uv sync --extra dev # install dependencies (runtime + pytest)
uv run knowledge-coach-server
By default the server listens on http://127.0.0.1:8000 and exposes the MCP Streamable HTTP endpoint at http://127.0.0.1:8000/mcp/.
Useful scripts:
uv run knowledge-coach-dev– run the server with reload in development mode.uv run knowledge-coach-inspector– launch MCP Inspector pointed at the local server.uv run knowledge-coach-inspector-config– generate aninspector.config.jsonfile that preconfigures the connection.
2. React web client
cd web
npm install
npm run dev
The dev server runs on http://localhost:5173. The client reads the MCP endpoint from VITE_MCP_BASE_URL (defaults to http://127.0.0.1:8000/mcp/). To override, create a .env.local file in web/:
VITE_MCP_BASE_URL=http://127.0.0.1:8001/mcp/
Running tests & checks
# From server/
uv run pytest
# From web/
npm run lint
npm run build
Demo flow
- Start the MCP server (
uv run knowledge-coach-server). - Optionally launch the inspector (
uv run knowledge-coach-inspector) to browse resources, prompts, and tools. - Start the React app (
npm run dev) and openhttp://localhost:5173. - Submit a prompt (e.g., “Design a training plan for sales engineers adopting MCP”) and select an experience level. The web client performs the MCP handshake, calls the
coach_resourcestool, and renders structured recommendations and next steps.
Troubleshooting
- Ensure the server is running before opening the React client; otherwise the client status will remain
idleor entererror. - For cross-origin calls the server allows
http://localhost:5173by default. If you access the client through a different origin, adjustKC_ALLOWED_ORIGINSinserver/.env. - If the inspector cannot connect, confirm the server is running and the URL matches
http://127.0.0.1:8000/mcp/.
Additional notes
- The MCP server ships with an in-memory knowledge base defined in
knowledge_base.yaml, plus resources, prompts, and a structured tool (coach_resources) that returnsCoachingResponsedata consumed by the UI. - React components are organised under
web/src/components/chat/and the MCP HTTP client lives inweb/src/services/mcpClient.ts.
Happy demoing!