Agent Shrine

Tools for AI agents

BreezeML exposes a set of safe, composable tools via the Model Context Protocol (MCP). Plug BreezeML into your agent and let it inspect data, train models, and deliver insights, while you stay in control.

View MCP Docs See Example Agent

A small forest shrine with a torii gate and hanging lantern among cedars

Tools · available MCP tools

inspect_data Profile dataset, summary stats, schema, quality	Data
compare Compare models across metrics and datasets	Model
train Train a model with specified data and params	Model
predict Generate predictions for new data	Inference
explain Explain predictions (SHAP, feature importance)	Inference
card Generate model card with key details	Governance
export Export model artifacts and metadata	Deployment
deploy Deploy model to a target environment	Deployment
save Save experiment, metrics, and artifacts	Project

Agent conversation (example)

Can you inspect the churn dataset and show me a summary?

tool_call: inspect_data { dataset: "churn_v1" }

tool_response: inspect_data

10,248

Rows

24

Columns

1.3%

Missing

churn

Target

14

Numeric

9

Categorical

1

Date

2

High card.

Train a logistic regression and compare it to a random forest.

tool_call: train { model: "logistic_regression" }

tool_call: train { model: "random_forest" }

tool_call: compare { models: [...] }

Start MCP server

UVPythonDocker

# Install
uv pip install breezeml[mcp]

# Start MCP server
breezeml mcp serve --host 127.0.0.1 --port 8787

MCP endpoint: http://127.0.0.1:8787/mcp

Connect your agent

Claude DesktopCursorCustom

{
  "mcpServers": {
    "breezeml": {
      "url": "http://127.0.0.1:8787/mcp"
    }
  }
}

Last stop: rest in the Zen Garden, and improve, every day.

Continue to the Zen Garden