Exact answers
about text.

An MCP server for the questions a language model answers by guessing. Counting, readability, diffs, regular expressions and hashes, computed and handed back.

Five tools. No API keys, no configuration, nothing to hold on to between calls.

A model reads tokens, not characters. Ask one how many letters are in a paragraph and it estimates, often badly. The same is true of diffing two files by eye, or predicting what a regex will match without running it. These are not hard problems. They are simply not the kind of problem a model is built to solve, so it should ask instead.

IFive tools

counting

count_text

Characters, words, lines, sentences, bytes. Visible characters reported apart from code points, because an accent is one and not two.

prose

score_readability

Flesch reading ease and grade level, with a plain reading of the number and the longest sentence named.

comparison

diff_texts

A unified diff, plus exact counts of lines added and removed and a similarity percentage.

patterns

test_regex

Runs the pattern. Every match with its position and capture groups; a broken pattern returns the error, not an exception.

digests

hash_text

md5, sha1, sha256 or sha512. Deterministic, and not something that can be worked out by reading.

IITry it

The counter above updates as you type, in the browser. Press Run and the same numbers come back from the server, computed by the functions the MCP tools call.

IIIConnect it

Claude Code, in one line:

# hosted claude mcp add --transport http vellum https://vellum-mcp.vercel.app/mcp

Claude Desktop, in claude_desktop_config.json:

{ "mcpServers": { "vellum": { "command": "python", "args": ["-m", "vellum.server"], "cwd": "/path/to/vellum-mcp" } } }

IVWhat it answers with

AskToolInstead of
How long is this really?count_textan estimate from token count
Is this readable?score_readabilityan impression
What changed?diff_textsreading both and describing
Does this pattern work?test_regexreasoning about it
What is the checksum?hash_textnothing, it cannot