MCP Server Integration

Overview

Published snapper / snapper-fmt binaries include an MCP (Model Context Protocol) server. The mcp Cargo feature is on by default.

Start it with:

snapper mcp

This launches the server on stdin/stdout using the MCP protocol.

Agents should call this server or the snapper CLI to apply semantic line breaks. Do not apply sembr.org or sembr/skills rewrites by hand.

From source with --no-default-features, rebuild with MCP:

cargo install snapper-fmt --features mcp

Configuration

MCP client configuration

Add to your MCP configuration:

{
  "mcpServers": {
    "snapper": {
      "command": "snapper",
      "args": ["mcp"]
    }
  }
}

Tools

format_text

Format text with semantic line breaks.

Parameters:

• text (string, required): Text to format

• format (string): Document format („org”, „latex”, „markdown”, „rst”, „plaintext”; default: „plaintext”)

• max_width (integer): Maximum line width (default: 0, unlimited)

• clause_breaks (boolean): Prefer soft breaks after independent-clause punctuation (same as CLI --clause-breaks; default: false). With max_width 0, always break at whitespace after the punctuation. With max_width greater than 0, overflowing sentences prefer those marks.

• range (object): Optional 1-indexed inclusive line range (start / end; same meaning as CLI --range)

• extra_abbreviations (array of strings): Additional abbreviations

Returns the formatted text.

detect_format

Detect document format from text content using heuristics.

Parameters:

• text (string, required): Text to analyze

Returns the detected format string.

check_formatting

Check text for semantic line break violations.

Parameters:

• text (string, required): Text to check

• format (string): Document format (default: „plaintext”)

• max_width (integer): Maximum line width; also the long threshold when set (default: 0)

• clause_breaks (boolean): Same two-mode contract as format_text (default: false)

Returns:

• would_reformat (boolean): identical to CLI --check (formatted output differs from input)

• passed (boolean): not would_reformat

• diagnostics (array): 1-indexed line / kind (fused / wrap / long) / excerpt

• violations (array): 1-indexed fused line numbers (compat)

split_sentences

Split text into individual sentences using Unicode-aware boundary detection.

Parameters:

• text (string, required): Text to split

Returns an array of sentence strings.