MCP Tools Lab

Seven local developer tools. One Python MCP server.

MCP Tools Lab brings everyday text and data utilities into a single MCP server. Format JSON, inspect CSV, compare text, generate hashes, convert timestamps, and inspect URLs from an MCP client or directly from Python. Every tool runs locally: no API keys, AI models, paid services, or outbound requests. Internet access is needed only to download dependencies during setup.

MCP (Model Context Protocol) lets a compatible application discover and call tools. This project exposes all seven through one local standard-input/output server. You can also call the Python functions directly, without an MCP client or model.

The seven tools

1. analyze_text — count characters, words, lines, approximate sentences, and the ten most frequent words.

2. process_json — validate, pretty-print, or minify JSON; optionally sort keys.

3. inspect_csv — show headers, data-row counts, missing values, and inconsistent row lengths.

4. hash_text — create SHA-256, SHA-512, or BLAKE2b hashes of UTF-8 text.

5. compare_text — generate a unified line diff and count added and removed lines.

6. convert_timestamp — convert Unix seconds and timezone-aware ISO 8601 dates.

7. inspect_url — split a URL into its components without visiting it.

Related MCP server: buildutilities-mcp

Quick start

Install Python 3.10 or newer and Git. Then:

git clone https://github.com/github-community-gitam/MCP-tools-lab.git
cd MCP-tools-lab
python -m venv .venv

Activate your environment:

# Windows PowerShell
.venv\Scripts\Activate.ps1

# macOS / Linux
source .venv/bin/activate

If Windows blocks activation, use .venv\Scripts\python.exe instead of python in the following commands; changing your execution policy is unnecessary. On systems where the command is python3, use that to create the environment.

python -m pip install .
python examples/try_tools.py

Connect an MCP client

Run the server with python -m mcp_tools_lab or mcp-tools-lab. It waits for MCP messages on standard input; a quiet terminal is normal. Stop a manually started server with Ctrl+C. Usually your MCP client starts it for you.

For clients that accept an mcpServers configuration, use this example and replace the executable with the absolute path to your virtual environment's Python:

{
  "mcpServers": {
    "mcp-tools-lab": {
      "command": "C:/path/to/MCP-tools-lab/.venv/Scripts/python.exe",
      "args": ["-m", "mcp_tools_lab"]
    }
  }
}

On macOS/Linux, the executable is /absolute/path/to/MCP-tools-lab/.venv/bin/python. Client configuration locations vary; use your client's local stdio server settings. The project uses the official Python MCP SDK's v1 API and bounds its dependency below v2 to avoid incompatible API changes.

Try tools with ordinary Python

from mcp_tools_lab.tools.json_tool import process_json
from mcp_tools_lab.tools.timestamps import convert_timestamp

print(process_json('{"b":2,"a":1}', sort_keys=True))
print(convert_timestamp("0"))
# {'unix_seconds': 0.0, 'iso_utc': '1970-01-01T00:00:00Z'}

See examples/try_tools.py for all seven functions and docs/tool-guide.md for inputs, results, and edge cases.

Project layout

src/mcp_tools_lab/
    server.py           # Registers the seven tools
    tools/              # One small Python module per tool
tests/                  # Tool tests and a real MCP round-trip test
examples/try_tools.py   # Run all tools without a model or MCP client
docs/tool-guide.md      # Input/output behavior

All application, example, and test code is Python. TOML and Markdown provide packaging configuration and documentation; no JavaScript or frontend is required. Tools accept supplied strings, do not read arbitrary files, and process inputs in memory. The server is designed for local use with small text inputs.

License

This project is released under the MIT License.

Contributors

• Kush Agarwal

• Driti Gudla