MCP Learning Console

An interactive lab project demonstrating a Model Context Protocol server and client through a polished Gradio dashboard.

Experiments included

Experiment 1 — Personal Assistant Memory

Persistent notes are stored in notes.db, a local SQLite database managed by SQLAlchemy and accessed only through MCP:

• save_note(content, tags) — create

• search_notes(query) and list_notes() — read

• update_note(note_id, content, tags) — update

• delete_note(note_id) — delete

Experiment 2 — Data Dashboard Connector

get_current_weather(location) retrieves current structured weather data from wttr.in. The UI displays temperature, condition, humidity, wind, and the raw MCP response.

The project also includes suggest_study(subject) for generating a short study path and interactive completion checklist.

Related MCP server: Custom MCP Server with RAG & Tools

SQLite database

database.py defines the SQLAlchemy Note model and creates C:\Users\licha\Downloads\mcp\notes.db automatically. SQLite is configured with check_same_thread=False and one session per MCP call so simultaneous Gradio requests are safe.

The first startup migrates valid records from the original notes.json, preserving IDs, content, tags, and timestamps. After a successful import the source is renamed to notes.json.backup; it is no longer read by the server. The backup can be opened for inspection, but notes.db is the only active memory store.

To inspect the database, use any SQLite browser or run:

python -c "from sqlalchemy import create_engine, text; e=create_engine('sqlite:///notes.db'); print(e.connect().execute(text('select id, content, tags from notes')).all())"

Install and run

Open Anaconda Prompt:

conda activate Gen_ai
cd /d C:\Users\licha\Downloads\mcp
python -m pip install -r requirements.txt
python app.py

The dashboard opens at http://127.0.0.1:7860. You can also double-click run_gen_ai.bat.

The command-line client is still available:

python client.py
python client.py --demo

LLM routing

The client uses a Groq/OpenAI-compatible chat endpoint when the following variables exist in the local .env file:

OPENAI_API_KEY=your_key
OPENAI_BASE_URL=https://api.groq.com/openai/v1
OPENAI_MODEL=openai/gpt-oss-20b

The real .env file is ignored by Git and must never be uploaded. Use .env.example as the safe configuration template. If the API is unavailable, the client automatically falls back to its deterministic local router.

Dashboard tabs

• AI Assistant: natural-language chat plus visible MCP tool selection, arguments, reason, router provider, and activity log

• Memory Notes: complete note CRUD interface with search and status feedback

• Weather Dashboard: live weather cards and structured tool output

• Study Planner: subject suggestions and progress checklist

Submission screenshots

Store screenshots in the screenshots folder. Recommended captures:

1. AI Assistant showing a successful Groq LLM decision and MCP tool trace.

2. Memory Notes showing saved notes and CRUD controls.

3. Weather Dashboard showing live weather cards.

4. Study Planner showing a generated learning path and checked progress.

Do not include the .env file or API key in screenshots or GitHub commits.

Project files

• app.py — interactive Gradio dashboard

• server.py — MCP server and tools

• client.py — reusable MCP client, LLM router, and CLI

• database.py — SQLite engine, SQLAlchemy model, and one-time migration

• notes.db — persistent local memory (generated and ignored by Git)

• notes.json.backup — preserved pre-migration JSON backup, if imported

• requirements.txt — pinned dependencies

• run_gen_ai.bat — Windows UI launcher

• run_cli.bat — Windows terminal-client launcher

References: MCP Python SDK, Groq OpenAI compatibility, and Gradio documentation.