Introduction
The Edumaths360 Practice Server is a small, working Model Context Protocol (MCP) server. It lets an AI assistant, such as Claude Desktop or Cursor, run practice in all three CAT sections, quantitative aptitude (QA), logical reasoning (DILR) and verbal ability (VARC), for CAT, other MBA entrance tests and similar competitive exams.
The key design rule is simple: the assistant should help the aspirant think, not think for them. In an exam hall there is no chatbot, so practice has to build the student's own method and speed. The server creates problems, marks answers, and gives step-by-step hints, but it never hands over the answer early.
:::tip What is MCP, in one sentence? MCP is an open standard that lets an AI application (the host) discover and call tools in another program (the server) through a common JSON-RPC message format. Think of it as a USB-C port for AI apps. Read How MCP works for the full picture. :::
A short walkthrough video (under 3 minutes) will appear here. It will show the server running in Claude Desktop: generating a problem, giving a hint, and marking an answer.
Placeholder: the video has not been recorded yet, so there is no link.
What the server offers
| Capability | Name | What it does |
|---|---|---|
| Tool | generate_problem | Creates a CAT-style problem for a topic and difficulty. Returns the question, never the answer. |
| Tool | check_answer | Marks an attempt and gives targeted feedback, for example "the gain % is worked out on the cost price". |
| Tool | get_hint | Gives a hint at level 1, 2 or 3. Hints never state the final answer. |
| Resource | syllabus://topics | Lists every topic with its exam section and area. |
| Resource template | syllabus://section/{section} | Lists the topics in one section: quant, reasoning or varc. |
| Prompt | socratic_tutor | A ready-made tutoring session that asks guiding questions. |
Topics, each at easy, medium and hard difficulty:
| Section | Area | Topics |
|---|---|---|
| Quant | Arithmetic | percentages, profit and loss, ratio and proportion, time-speed-distance, time and work, simple and compound interest |
| Quant | Algebra | linear equations, quadratic equations |
| Quant | Number system | HCF and LCM, units digits, counting factors |
| Reasoning | Logical reasoning | number series |
| VARC | Verbal ability | para jumbles, odd sentence out, para summary (original passages written for this project) |
Who these docs are for
- Developers who want to learn MCP by reading a small, complete, tested server.
- Coaching teachers and EdTech builders who want an AI assistant that supports exam preparation instead of replacing the student's own thinking.
- Anyone evaluating MCP, who wants real request and response examples rather than pseudo-code.
How these docs are organised
The docs follow the Diátaxis framework, which splits documentation by what the reader needs at that moment:
| Section | Use it when you want to… | Start with |
|---|---|---|
| Tutorials | learn by doing, step by step | Build your first MCP server |
| How-to guides | complete a specific task | Add a new tool |
| Reference | look up exact facts: schemas, errors, payloads | Tools reference |
| Explanation | understand the design and the reasons behind it | Architecture |
Quickstart
You need Python 3.10 or later and uv, a fast Python package manager.
git clone https://github.com/Creator-hub24/edumaths360-mcp-docs.git
cd edumaths360-mcp-docs/server
uv sync --extra dev # create .venv and install the server + test tools
uv run pytest -q # run the test suite
uv run python ../snippets/client.py # watch one full practice round
The last command starts the server, connects a Python client to it, and prints a practice round. The output is shown in Connect from a Python client.
Accuracy
Every code sample on this site is copied from a file in the repository, and the CI pipeline runs those files on every change. The JSON-RPC examples were captured from the running server by snippets/capture_payloads.py. The server is built with the official MCP Python SDK version 2.x.