TransformerBee.MCP
This is a simple PoC of a Model Context Protocol (MCP) server for transformer.bee, written in Python.
Under the hood it uses python-mcp and transformerbeeclient.py.
Installation
You can install the MCP server as Python package or pull the Docker image.
Install as Python Package
or if you are using pip:
Install as Docker Image
Start the Server via CLI
Python
_The package ships a simple CLI argument to start the server. In a terminal inside the virtual environment in which you installed the package (here , call:
Docker
(For the environment variables -e ..., see below or the transformerbeeclient.py docs.)
Register MCP Server in Claude Desktop
If you checked out this repository
If you installed the package via pip/uv
Modify your claude_desktop_config.json (that can be found in Claude Desktop menu via "Datei > Einstellungen > Entwickler > Konfiguration bearbeiten"):
where C:\github\MyProject\.myvenv is the path to your virtual environment where you installed the package and localhost:5021 exposes transformer.bee running in a docker container.
Alternatively, if you haven't configured this handy CLI command
https://github.com/Hochfrequenz/TransformerBee.mcp/blob/c0898769670469df13f23b57a55fe4b71ed9795b/pyproject.toml#L101-L102
you can just call python with non-empty args.
Note that this package marks uv as a dev-dependency, so you might need to install it pip install transformerbeempc[dev] in your virtual environment as well as a lot of MCP tooling assumes you have uv installed.
For details about the environment variables and/or starting transformer.bee locally, check transformerbeeclient.py docs.
If you installed the package via Docker
I'm aware, that using the --network host option is a bit hacky and not best practice.