What it does
glin is an MCP server that lets an AI agent train an Explainable Boosting Machine classifier on a CSV or Excel file and then use it. Each prediction comes with an exact breakdown of the features behind it, taken from the model's additive structure rather than a SHAP or LIME approximation. The agent tests the model on rows it did not train on and reports how good it is. It can also tune models, explain what drives an outcome, and list or delete saved models.
Use it when you want to classify rows, such as sales leads or deals, and need the reasons for each result. You describe the task in plain language to an agent such as Claude Code, Claude Desktop or Cursor, so no code is needed. Saved models remain available in later chats. The server runs on your machine, and your data and models stay there.
Running it requires uv, which starts glin through uvx from the glin-ml package on PyPI. The Claude Code plugin adds the server and a skill that guides the agent, and includes Excel support. Other clients need a config entry for the server. The first run downloads glin and its libraries, which can take about a minute.
Limitations and requirements
Requires uv to be installed. The first start can take about one minute. Some clients, including Claude Code, stop a server that does not start within 30 seconds, so the first run should be done before adding it to a client. Claude Desktop may need the full path to uvx in its config if it cannot find the command. Clients must be restarted after config changes; on macOS, closing the Claude Desktop window is not enough.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
glin
Exactly explainable classifiers for agents: train, tune, predict, explain on CSV or Excel.
v0.1.5As shown on registry.modelcontextprotocol.io when we last read it