io.github.nexus-api-lab/textops-mcp

Compute an exact line-based unified diff between two texts. Use instead of writing diffs yourself — LLM-authored diffs have wrong line numbers and miscounted hunks.

Apply a unified diff to a text, with context verification and small-offset fuzzing. Reports per-hunk success so failures are explicit rather than silent.

Execute a JavaScript regular expression against a text and return real matches with indices and capture groups. Use instead of predicting regex behavior — LLM regex predictions are unreliable.

Count text length EXACTLY: graphemes (user-perceived characters, correct for emoji/CJK/combining marks), words, lines, Unicode code points, UTF-16 units, UTF-8 bytes. Use whenever "how many characters" matters.

Validate JSON and return it canonicalized (recursively sorted object keys) in compact and pretty forms. Useful for stable comparison and exact validation with error details.