What it does
DataClean provides four paid tools and two free helpers. profile_table reports the encoding, delimiter, header row, and per-column type, number locale, date order, nulls and duplicates, plus a suggested schema. clean_table returns a typed copy as CSV, TSV, JSON, JSONL or .xlsx and lists every value it could not convert, with its source row. validate_schema checks each row against a schema and returns errors with row numbers. normalize_values processes up to 10,000 values taken from a document or another tool. pricing and credits_balance are free.
It is aimed at tables that language models misread, such as amounts like 1.250,50 € or dates like 03/04/2026 where day and month are ambiguous, IDs with leading zeros, and files with NUL bytes or mojibake. Values are converted only when the column or a supplied parameter settles their meaning; otherwise they are kept and flagged. Text cells that begin with spreadsheet formula characters are prefixed in CSV output.
Each paid call consumes credits. Invalid input is refused before any charge, and a failure after charging is refunded. Setup involves buying credits to obtain a license key, then adding the server to an MCP client; Claude Desktop can also install a bundled package, and a hosted mode reads the license from a bearer header on each request. Input and output size and row limits are set through environment variables. In Claude, ChatGPT, VS Code and Goose, paid tools display an interactive card with the balance or result summary.
Limitations and requirements
Paid tools require credits and a KeyFlint license key (LIC-...). Production use requires a license from the Licensor under the Business Source License 1.1. Default limits are 20 MB input, 200,000 rows and 15 MB output, adjustable by environment variables. Schema regexes with nested quantifiers are refused. Interactive cards are shown only in the listed clients; other clients receive JSON.
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 15 Nov 2026 · its page on the source
DataClean
Profile, clean and validate CSV/Excel data: schemas, dates, numbers, duplicates. Never guesses.
v0.2.1As shown on registry.modelcontextprotocol.io when we last read it