DataMotto

Every notebook assigns an agent to the job in front of you. The Data Prep Agent works on structure and cleanliness, and the Analysis Agent takes over once the table is stable, running exploratory charts, summaries and simple models. Splitting the work keeps each agent focused, and it gives you a clear place to check the reasoning before the analysis moves on.

Upload a CSV or Excel file and DataMotto detects data types, renames sloppy column headers, drops empty rows and columns, and removes columns that hold one repeated value. That last step matters more than it sounds: a constant column tells a model nothing and often hides a broken export. You review the proposed changes rather than accept a silent rewrite, which keeps the original file honest.

The data quality score grades how ready a dataset is for the question you asked, not how tidy it looks in general. It lists the issues it found, checks the data against your stated objective and suggests enrichment where a column is too sparse to support a conclusion. Read it as a triage signal: a low score means slow down, not that the file is worthless.

Pick a column, highlight an example of what you want and describe it in a sentence. The agent writes the extraction code and drops the result into a new column, which is faster than building a regex by hand and easier to audit because the code stays visible. It handles order references, product names and address fragments well, and it struggles when one column mixes several formats with no consistent marker.

Instead of exporting a spreadsheet every week, you can connect a Notion database, HubSpot records or Stripe data and let the notebook retrieve it. The connection converts records into a dataframe, so the same cleaning and analysis steps apply. It suits recurring reporting, and it does mean your analysis is only as fresh as the last sync.

Each notebook page can run Python, R and Julia in separate sessions, with a long list of statistics packages preinstalled, and you can install anything missing in a reproducible environment. Variables don't cross sessions, so a dataframe built in Python isn't visible to R. That's a real constraint for mixed workflows, and it's also the reason a notebook stays predictable when you hand it to someone else.