MCP Toolkit Server

A small, production-shaped Model Context Protocol (MCP) server that gives any MCP-compatible AI client — Claude Desktop, Claude Code, or a custom agent — a set of safe, well-defined tools it can call to reach real data and take real actions.

It's built to be read in five minutes and extended in ten. If you want an AI agent that can actually do things in your systems instead of just talking, this is the pattern that makes it happen.

---

What it does

The server exposes five tools over the MCP protocol:

Tool	What it does
fetch_url(url)	Fetches a web page and returns clean, readable text
extract_metadata(text)	Pulls structured fields (emails, phones, URLs, dates) out of messy text
csv_summary(path)	Loads a CSV and reports its shape and per-column stats
now()	Returns the current UTC timestamp
word_count(text)	Counts words, characters (with and without spaces), and lines

Each tool is a plain Python function with a docstring. The MCP layer turns those docstrings into the schema the AI model reads to decide when and how to call each tool. Add a new tool by writing a new decorated function — that's the whole extension model.

---

Quickstart

git clone <this-repo>
cd mcp-toolkit-server
pip install -r requirements.txt
python src/server.py

The server runs over stdio, which is what Claude Desktop and Claude Code expect for a local MCP server.

Connect it to Claude Desktop

Add this to your Claude Desktop MCP config (claude_desktop_config.json):

{
  "mcpServers": {
    "toolkit": {
      "command": "python",
      "args": ["/absolute/path/to/mcp-toolkit-server/src/server.py"]
    }
  }
}

Restart Claude Desktop, and the four tools appear. Ask it to "fetch example.com and pull out any emails" and watch it chain fetch_url → extract_metadata on its own.

---

Why this pattern matters

Most "AI integration" work is exactly this: taking a model that can reason and giving it safe, scoped tools to act on real systems — your CRM, your files, your APIs, your database. MCP is the emerging standard for doing that cleanly, and this repo is a minimal, honest example of the shape.

The same structure scales to production tools: a Shopify tool that reads order data, a database tool with read-only queries, an internal-API tool with proper auth. The tools change; the pattern doesn't.

---

Built by

I build AI-agent and automation systems — Claude Code / MCP integrations, LLM-into-workflow tooling, scrapers, and scripts that kill repetitive work. If you need an agent wired into your real systems, that's the work I do.

• Upwork: its_weli on Upwork
• Contact: hello@weli.build

---

License

MIT — use it, fork it, ship it.