claude_tokenizer_peek_columns – 1 MCP server offers this tool
What it usually takes
config– in 1 of 1dataset_id– in 1 of 1split– in 1 of 1
No server marks it as only reading: treat it as able to change data.
From each publisher's own tool list. We never call a tool.
Servers
| Server | Inputs | Read-only? |
|---|---|---|
Claude Tokenizer co.huggingface.sanderland/claude-tokenizer Fill the column picker, ranked by mean word count over the first few rows. Not the first row's longest string: row 0 is often a header or canary whose fields are all the same length, and not length either — an id like `test-en-news.3585` is long and is not text. Words per row separates prose from identifiers whatever their length. | config,dataset_id,split | Not stated |