use case
Give messy CSV exports to your agent without the parsing tax.
Frenchie converts CSV files into Markdown tables so your agent can spot bad rows, explain fields, and draft lightweight reports.
the problem
Why this is a pain.
A product export lands as CSV. It has quoted commas, blank fields, long rows, and 8,000 records. You do not need a warehouse job. You need your agent to read enough of it to understand the shape and help clean it up.
Raw CSV pasted into chat wastes context and breaks on edge cases. A script can parse it, but then you are maintaining another tiny data tool. Frenchie handles the file-to-Markdown step and stops there.
the workflow
How Frenchie handles it.
- 1
Install Frenchie once so the MCP tools and skills land together.
- 2
Ask your agent to extract the CSV and name the cleanup you want — missing emails, duplicate IDs, suspicious totals, inconsistent status values.
- 3
Frenchie parses the CSV with standard quoting rules and returns agent-friendly Markdown tables. Large files may be chunked, with row samples and truncation notes emitted per chunk.
- 4
Your agent reviews the Markdown table and writes the cleanup plan, report, or transformation script.
example
What it looks like in practice.
Your prompt
Extract ./exports/customers.csv with Frenchie. Tell me which columns look inconsistent and write a cleanup checklist.
Agent response
CSV extracted — 0.5 credits, Markdown tables saved to .frenchie/customers/result.md. Large-file samples and truncation notes are shown per chunk when needed. Cleanup checklist: - Normalize `plan` values: free, Free, trial, Trial. - Split `name` into first/last only after checking 41 single-token names. - Validate emails missing `@`. - Review rows where `mrr` is blank but `status` is active.
What lands in your workspace
$frenchie extract forecast.xlsx
# Workbook: forecast.xlsx ## Sheet: Summary | Metric | Q3 | Q4 | | ------ | -- | -- | | ARR | 4.2M | 4.9M | | CAC | 820 | 780 | ## Sheet: Assumptions - Expansion motion starts in May - Partner channel adds 12 deals
tips
Things worth knowing.
- CSV extraction is a flat 0.5 credit per file.
- Frenchie renders a sample-sized Markdown table, not a full analytics warehouse.
- Ask your agent for a script after extraction if you need to clean the original file.
questions
Common questions.
Does Frenchie process every CSV row?
It processes the CSV into agent-friendly Markdown tables. Large files may be chunked, with samples and truncation notes emitted per chunk.
Can this replace a database?
No. Use a database or warehouse for large-scale analysis. Frenchie is for agent-readable file extraction.
How much does CSV extraction cost?
0.5 credit per CSV file.
Try it with a real Office file of yours.
100 free credits on signup. No card. Drop a DOCX, XLSX, CSV, TSV, or PPTX file from your own workflow and see the Markdown your agent gets back.