diff --git a/sample_data/description.md b/sample_data/description.md
index 813d075..ad81314 100644
--- a/sample_data/description.md
+++ b/sample_data/description.md
@@ -1,12 +1,73 @@
-# Dataset description
+# Sample datasets
 
-## classification dataset
-- titanic.csv
+This folder contains small CSV files used in the unit tests and the examples. They are handy for quickly trying out the library without having to download any external data.
 
+## titanic.csv
 
-## regression dataset
-- house-prices.csv
+A subset of the famous Titanic passenger list. Each row represents a passenger and the `Survived` column is a binary label indicating whether that person survived the disaster. You can use it to build a simple classification model.
 
+Columns include:
 
-## clustering dataset
-- iris.csv
+- `PassengerId` – unique passenger identifier
+- `Survived` – target label (0 = did not survive, 1 = survived)
+- `Pclass` – ticket class
+- `Name`, `Sex`, `Age` – basic personal information
+- `SibSp`, `Parch` – number of siblings/spouses or parents/children aboard
+- `Ticket`, `Fare`, `Cabin`, `Embarked` – ticket details
+
+## house-prices.csv
+
+Synthetic house price data for regression tasks. The goal is to predict the `Price` column based on features such as the size of the house and the neighbourhood.
+
+Columns include:
+
+- `Home` – row identifier
+- `Price` – target value (house price)
+- `SqFt`, `Bedrooms`, `Bathrooms` – property characteristics
+- `Offers` – number of purchase offers received
+- `Brick` – whether the house is made of brick
+- `Neighborhood` – area of the city (`East`, `West`, `North` ...)
+
+## iris.csv
+
+Classic Iris flower measurements. It can be used for clustering or classification. The `variety` column gives the species of the plant.
+
+Columns include:
+
+- `sepal.length`, `sepal.width`, `petal.length`, `petal.width`
+- `variety` – Setosa, Versicolor or Virginica
+
+## Using the data with `faster_llm`
+
+Each dataset can be loaded with `pandas` and passed into the high level models provided by the library. Metrics are calculated automatically and can optionally be forwarded to an MCP compatible service.
+
+Example using the Titanic data:
+
+```python
+import pandas as pd
+from sklearn.linear_model import LogisticRegression
+from faster_llm.ML.classification import Model
+
+# load the CSV
+frame = pd.read_csv("./sample_data/titanic.csv", sep="\t")
+X = frame.drop(columns=["Survived"])
+y = frame["Survived"]
+
+# train and send metrics to an MCP server
+clf = Model(
+    LogisticRegression(max_iter=200),
+    X,
+    y,
+    send_to_llm_flag=True,
+    server_url="http://localhost:8000",
+)
+```
+
+For regression you can use `house-prices.csv` with `faster_llm.ML.regression.Model` in the same way. Clustering with `iris.csv` can be performed via `faster_llm.ML.clasterization.ClusterModel`.
+
+To send arbitrary results to an MCP server you may also call `send_to_llm` directly:
+
+```python
+from faster_llm.LLM import send_to_llm
+send_to_llm("Training complete", server_url="http://localhost:8000")
+```