MCP Server

The malarky-mcp package is an MCP (Model Context Protocol) server that lets LLMs generate nonsense text, apply morphology operations, and validate lexicons – directly as tool calls.

It works with Claude Desktop, Claude Code, Cursor, and any other MCP-compatible client.

Installation

npm install -g malarky-mcp

Or run without installing:

npx malarky-mcp

Setup

Claude Desktop

Add to your Claude Desktop config file (claude_desktop_config.json):

{
  "mcpServers": {
    "malarky": {
      "command": "npx",
      "args": ["malarky-mcp"]
    }
  }
}

Claude Code

claude mcp add malarky -- npx malarky-mcp

Cursor

Add to your Cursor MCP settings (.cursor/mcp.json):

{
  "mcpServers": {
    "malarky": {
      "command": "npx",
      "args": ["malarky-mcp"]
    }
  }
}

Tools

The server exposes eight tools. All parameters are optional unless marked required.

---

generate_sentence

Generate one or more sentences of syntactically plausible English nonsense.

Parameter	Type	Description
seed	number	RNG seed for deterministic output
type	string	Sentence structure (see list_sentence_types)
count	number	Number of sentences to generate (1–50, default: 1)
minWords	number	Minimum words per sentence
maxWords	number	Maximum words per sentence
hints	string	Comma-separated tags (e.g. domain:tech,register:formal)
transforms	string	Comma-separated transform IDs (e.g. pigLatin,leet)
archetype	string	Archetype name to activate from the lexicon
lexicon	string	Lexicon JSON string for domain-specific vocabulary
lexiconPath	string	Absolute path to a lexicon JSON file on disk
config	string	Generator config JSON (see list_config)

Example prompt: “Generate 3 sentences in Pig Latin with seed 42”

The LLM calls generate_sentence with count: 3, seed: 42, transforms: "pigLatin".

---

generate_paragraph

Generate one or more paragraphs.

Parameter	Type	Description
seed	number	RNG seed for deterministic output
count	number	Number of paragraphs (1–20, default: 1)
sentences	number	Fixed number of sentences per paragraph
minSentences	number	Minimum sentences per paragraph
maxSentences	number	Maximum sentences per paragraph
hints	string	Comma-separated tags
transforms	string	Comma-separated transform IDs
archetype	string	Archetype name to activate from the lexicon
lexicon	string	Lexicon JSON string for custom vocabulary
lexiconPath	string	Absolute path to a lexicon JSON file on disk
config	string	Generator config JSON (see list_config)

---

generate_text

Generate a text block (multiple paragraphs).

Parameter	Type	Description
seed	number	RNG seed for deterministic output
paragraphs	number	Fixed number of paragraphs
minParagraphs	number	Minimum paragraphs
maxParagraphs	number	Maximum paragraphs
hints	string	Comma-separated tags
transforms	string	Comma-separated transform IDs
archetype	string	Archetype name to activate from the lexicon
lexicon	string	Lexicon JSON string for custom vocabulary
lexiconPath	string	Absolute path to a lexicon JSON file on disk
config	string	Generator config JSON (see list_config)

---

morphology

Apply English morphology operations to a word – pluralize, conjugate verbs, get the indefinite article, and more.

Parameter	Type		Description
word	string	required	The word to transform
operation	string	required	The operation to apply

Operations:

Operation	Example input	Example output
pluralize	synergy	synergies
singularize	stakeholders	stakeholder
pastTense	go	went
pastParticiple	run	run
presentParticiple	leverage	leveraging
thirdPerson	do	does
indefiniteArticle	hour	an

---

list_transforms

List all available output transforms. No parameters.

Returns the 10 built-in transforms: pigLatin, ubbiDubbi, leet, uwu, pirate, redact, emoji, mockCase, reverseWords, bizJargon.

See Output Transforms for detailed descriptions of each.

---

list_sentence_types

List all available sentence structure types with descriptions. No parameters.

Returns the six types: simpleDeclarative, compound, introAdverbial, subordinate, interjection, question.

See Guides > Sentence Types for examples of each.

---

list_config

List all generator configuration options with their types and default values. No parameters.

Use this to discover what can be passed in the config parameter of the generation tools. See Configuration for full details on each option.

---

validate_lexicon

Validate a lexicon JSON string or file and report any errors or warnings. Provide either json or lexiconPath.

Parameter	Type	Description
json	string	The lexicon JSON string to validate
lexiconPath	string	Absolute path to a lexicon JSON file on disk

Returns validation status, a list of errors (if any), and warnings.

See Lexicons for the full schema reference and how to build custom lexicons.

Using custom lexicons

There are two ways to load a lexicon into any generation tool:

1. By file path – pass lexiconPath with an absolute path to a .json file on disk. This is the easiest option when you already have a lexicon file.
2. Inline JSON – pass lexicon with the full JSON string. Useful for one-off or dynamically constructed lexicons.

If both are provided, lexiconPath takes precedence.

Example prompt: “Generate a startup-themed paragraph using the lexicon at /home/me/lexicons/startup.json”

The LLM calls generate_paragraph with lexiconPath: "/home/me/lexicons/startup.json" and archetype: "startup".

Inline lexicon example

For quick, ad-hoc lexicons, an LLM can pass the JSON directly:

{
  "id": "lexicon.startup",
  "language": "en",
  "termSets": {
    "noun.startup": {
      "pos": "noun",
      "tags": ["domain:startup"],
      "terms": [
        { "value": "disruptor" },
        { "value": "unicorn" },
        { "value": "pivot" },
        { "value": "runway" }
      ]
    },
    "verb.startup": {
      "pos": "verb",
      "tags": ["domain:startup"],
      "terms": [
        { "value": "disrupt" },
        { "value": "scale" },
        { "value": "pivot" },
        { "value": "iterate" }
      ]
    }
  },
  "archetypes": {
    "startup": {
      "tags": ["domain:startup"]
    }
  }
}

Generator configuration

All generation tools accept a config parameter – a JSON string with any GeneratorConfig overrides. This lets you tune sentence structure rates, word counts, complexity limits, and more.

Example prompt: “Generate a paragraph with lots of questions and long sentences”

The LLM calls generate_paragraph with:

{
  "config": "{\"questionRate\":0.8,\"minWordsPerSentence\":15,\"maxWordsPerSentence\":30}"
}

Common config options:

Option	Type	Default	Description
minWordsPerSentence	number	5	Minimum words per sentence
maxWordsPerSentence	number	25	Maximum words per sentence
questionRate	number (0-1)	0.10	Rate of question sentences
compoundRate	number (0-1)	0.15	Rate of compound sentences
subordinateClauseRate	number (0-1)	0.15	Rate of subordinate clauses
maxAdjectivesPerNoun	number	2	Maximum adjectives before a noun
maxAdverbsPerVerb	number	1	Maximum adverbs per verb phrase
maxPPChain	number	2	Maximum prepositional phrase chains
minSentencesPerParagraph	number	2	Minimum sentences per paragraph
maxSentencesPerParagraph	number	7	Maximum sentences per paragraph
sentenceTypeWeights	object	–	Weights for each sentence type

Call list_config to see the full list with defaults.

Deterministic output

Pass the same seed value to any generation tool to get reproducible results. This is useful when an LLM needs consistent placeholder text across multiple calls.