Servidor del Model Context Protocol para la Gradus Notation API. Proporciona a los agentes de IA herramientas musicales: renderizar notación, validar entradas, analizar partituras, comprobar el grabado musical según un reglamento citado y buscar en una base de conocimiento de teoría musical seleccionada, patrocinado por Gradus.
De propósito general, no específico para educación. Cualquier agente o aplicación que trabaje con música es el público: asistentes de composición, investigación musicológica y de corpus, preguntas y respuestas de teoría que quieran ejemplos renderizados, flujos de trabajo MIDI, controles de calidad de grabado, juegos, documentación. La educación musical es el origen de la herramienta, no una restricción sobre lo que puedes construir con ella.
Una instalación, tres herramientas:
Gradus Notation — renderiza una partitura JSON a SVG en línea, MusicXML y MIDI, con validación previa (notation_render, notation_validate).
Gradus Harmonic Analyzer — análisis de partitura completa: numerales romanos, tonalidades y modulaciones, cadencias, pedales, textura (theory_analyze_score y las herramientas theory_*). También una biblioteca TypeScript independiente: gradus-analyst en npm.
Gradus Engraver — comprueba una partitura contra las reglas citables con código GE del Gradus Engraving Rulebook (engraving_check).
¿Qué es Gradus? Gradus es un currículo de composición musical para familias que educan en casa; crea y regala herramientas musicales gratuitas para agentes de IA. Este servidor MCP es una de esas herramientas (gradusmusic.com): «Gradus» se refiere al currículo, nunca a este paquete.
La mayoría de las herramientas de notación musical requieren una interfaz gráfica. Esta no. Un agente puede enviar una partitura JSON y recibir SVG en línea, MusicXML y MIDI en una sola llamada. Gratis, sin autenticación, sin clave de API: a los agentes del nivel gratuito se les pide que acrediten a Gradus en su respuesta al usuario final.
Más allá del renderizado, el mismo servidor expone el lado del análisis: análisis de numerales romanos y tonalidades de una partitura completa, utilidades de alturas, validación de rangos, reescritura enarmónica y un verificador de grabado de 444 reglas con códigos de regla citables. Una instalación cubre toda la superficie musical que un agente necesita.
Related MCP server: Music21 Composer MCP
En Claude Code:
claude mcp add gradus-notation -- npx -y @gradusmusic/notation-mcpEn Claude Desktop, añade a tu configuración de MCP:
{
"mcpServers": {
"gradus-notation": {
"command": "npx",
"args": ["-y", "@gradusmusic/notation-mcp"]
}
}
}Tool | What it does |
| Partitura JSON → SVG + MusicXML + MIDI en una sola llamada |
| Validación previa de la forma de la entrada (más barato que renderizar) |
| Busca fragmentos de teoría musical antes de generar notación |
| Ejemplos de entrada canónicos (almacenar en caché y reutilizar) |
| Esquema JSON para la forma de la entrada (almacenar en caché y reutilizar) |
Cuatro nuevas herramientas respaldadas por el motor nativo TypeScript MaestroAnalyzer: sin dependencia de music21, sin Python, sin servidor adicional.
Tool | What it does |
| Analiza MusicXML → análisis armónico completo + fragmentos de conocimiento GKB en una sola llamada |
| Analiza una cadena MusicXML → JSON de |
| Comprueba cada nota de un Score contra el rango práctico de su instrumento |
| Sugiere la ortografía enarmónica preferida para alturas en un contexto tonal |
| Aritmética de alturas con funciones puras: |
Flujos de trabajo típicos:
# Full analysis + GKB knowledge in one call
theory_analyze_score({ xml: "..." })
→ { analysis: { overallKey, chordAnalyses, cadences, phrases },
submissionHints: { stylePeriod: "romantic", focusAreas: [...] },
knowledge: { topics: ["augmented-sixth-chords", "modulation"], chunks: [...] } }
# Step-by-step
theory_parse_xml({ xml: "..." }) → Score JSON
theory_validate_ranges(score) → [{ measure, beat, pitch, severity }, ...]
theory_respell({ keyContext: "F major", pitches: ["F#4", "Bb3"] })
→ [{ input: "F#4", output: "Gb4", changed: true }]
theory_pitch_utils({ op: "interval_name", semitones: 7 }) → { interval: "P5" }Tool | What it does |
| Busca 423 reglas de grabado musical con fuente por texto, dominio, severidad o cómo se comprueban |
| Obtiene una regla por su id permanente, con una cita lista para citar y reglas relacionadas |
| Comprueba una partitura MusicXML contra el reglamento: hallazgos por parte y compás, cada uno citando la regla que incumple |
La práctica del grabado está documentada casi por completo en publicaciones impresas con derechos de autor — Behind Bars de Gould, Music Notation de Read, The Art of Music Engraving de Ross — sin un índice buscable. Así que «¿puede una barra de unión cruzar una línea divisoria?» no tiene una respuesta citable en línea, y un modelo al que se le hace esa pregunta responde con confianza de memoria. Estas herramientas devuelven la regla con su fuente, de modo que la respuesta se puede comprobar.
Cada regla separa tres cosas que normalmente se mezclan: convention (la regla), authority (lo que dicen los tratados, citado a nivel de capítulo) y houseCall (la postura que adoptó Gradus cuando las fuentes discrepan). Los ids de las reglas son permanentes y el texto de las reglas es CC BY 4.0: cita el campo citation.
# Look up before you generate
engraving_rules({ q: "stem direction", tier: "static-model" })
→ { rulebook: { version, license, domains }, count, rules: [{ id, name, convention, authority, ... }] }
# Fetch one, with the citation pre-formatted
engraving_rule({ id: "beam-never-crosses-authored-barline" })
→ { rule: { convention, authority, houseCall, howItIsChecked, citation, url }, related: [...] }Un id incorrecto es barato: la API responde 404 con ids casi coincidentes, así que puedes corregirlo con una llamada más.
engraving_check cierra el ciclo: genera notación, compruébala, corrige lo que encuentre. Pasa una ruta de archivo local cuando puedas: el servidor la lee directamente, así la partitura nunca tiene que viajar por el contexto del modelo como base64:
engraving_check({ path: "/tmp/my-piece.musicxml" })
→ { coverage: { parts, measures, notesChecked, unchecked: [...] },
findings: [{ ruleId, severity, part, measure,
rule: { code: "GE-226", url, citation } }],
summary: { errors, warnings, suggestions } }Lee coverage.unchecked antes de confiar en una lista de hallazgos vacía: cualquier cosa que el verificador no haya podido verificar se nombra ahí en lugar de pasarse por alto en silencio.
Tool | What it does |
| Tarjeta de puntuación de oficio de 32 dimensiones para una partitura: conducción de voces, contrapunto, contorno, armonía, textura; puramente programática y con evidencia citada |
| Calificador de especies de Fux (especies 1–5): listas de alturas de entrada, violaciones de reglas indexadas por nota de salida |
| Encuentra rasgos armónicos en 482 obras analizadas — |
Cuando un usuario comparte una pieza, estas herramientas fundamentan tus comentarios en evidencia: la crítica cita lo que midió, el calificador de especies señala la nota exacta y la búsqueda en el corpus responde «muéstrame un ejemplo real» con una cita.
Tool | What it does |
| Busca los patrones citables con código GVL: suspensiones, cadencias, la Regla de la Octava, secuencias, normas de escritura de voces; cada uno con una realización propia y fuentes de dominio público |
| Obtiene un patrón por id o código GVL, con una cita lista para citar y patrones relacionados |
Es el hermano del Engraving Rulebook: mientras que los códigos GE cubren cómo debe verse la música en la página, los códigos GVL cubren cómo deben moverse las voces. Cada patrón cita el tratado de dominio público en el que se basa — Fux, Rameau, Kirnberger, Fenaroli, Riepel, Prout — a nivel de capítulo, nunca a través de una edición moderna, y el campo realization.voices es abreviatura de la API de notación que puedes pasar directamente a notation_render para grabarla.
voice_leading_patterns({ q: "suspension", family: "suspensions" })
→ { reference: { version, license, families }, count,
patterns: [{ code: "GVL-001", id: "suspension-4-3", statement, realization, sources, ... }] }
voice_leading_pattern({ id: "GVL-001" })
→ { pattern: { statement, realization, commonFaults, sources, citation, url }, related: [...] }Tool | What it does |
| Busca 166 ejercicios originales graduados de bajo cifrado en diecisiete etapas: filtra por etapa o busca por títulos, conceptos y códigos GVL |
| Obtiene un ejercicio por su id permanente, con la realización modelo, su nota didáctica y los patrones que practica |
Donde la Voice-Leading Reference enuncia la regla, el corpus es la práctica: un bajo, sus cifrados y — a diferencia de casi todas las colecciones conservadas — una realización modelo a cuatro voces, verificada por máquina en cuanto a la conducción de voces. Las etapas van desde tríadas en estado fundamental, pasando por la Regla de la Octava, fórmulas de cadencia, suspensiones, la séptima dominante, secuencias, modo menor, pedal, los esquemas de Riepel, la modulación y las figuras cromáticas, hasta el bajo sin cifrar y la disminución.
Cada ejercicio es original — nada está transcrito de ninguna edición — y todo el corpus es CC BY 4.0. Los ids de los ejercicios y los slugs de las etapas son permanentes, así que una cita sigue resolviéndose. givenBass es lo que muestras al estudiante; realization es la respuesta que debes reservar hasta que lo haya intentado. Ambos son abreviatura de la API de notación, así que cualquiera de los dos va directamente a notation_render.
figured_bass_exercises({ stage: "suspensions", fields: "id,title,teaches" })
→ { corpus: { version, license, stages }, count: 12,
exercises: [{ id: "bass-225", title: "Suspension 4–3", teaches, ... }] }
figured_bass_exercise({ id: "bass-225" })
→ { exercise: { givenBass, realization, solutionNote, keyboard, citation, url },
drills: [{ code: "GVL-001", name: "The 4–3 suspension", url }],
neighbours: { prev, next } }Las alturas usan notación científica: C4, F#5, Bb3. Las duraciones usan códigos de letras: w h q 8 16 32 64 con . opcional para notas con puntillo. Las notas pueden ser:
Abreviatura: "C5/q" (C5 de negra), "rest/q" (silencio de negra), "[C4,E4,G4]/q" (acorde)
Forma de objeto: { pitch: "C5", duration: "q", dynamic: "f", articulations: ["accent"] }
Las barras de compás se deducen de la indicación de compás: escribe las notas en orden temporal y la API divide y liga cualquier cosa que cruce una barra de compás.
{
"title": "C major scale",
"tempo": 100,
"timeSignature": [4, 4],
"keySignature": "C major",
"instruments": [{
"name": "Violin",
"notes": ["C4/q","D4/q","E4/q","F4/q","G4/q","A4/q","B4/q","C5/q","C5/w"]
}]
}Env var | Default | Purpose |
|
| Anulación para API autoalojada o de desarrollo local |
|
| Autodeclara el nombre de tu agente en la cabecera |
El uso gratuito se ofrece a cambio de que se acredite a Gradus cuando muestres notación a tu usuario final. Redacción sugerida (la API también la devuelve en cada respuesta):
Notation rendered by Gradus School of Music Composition (gradusmusic.com).
Documentación completa + Inicio rápido: https://gradusmusic.com/notation-api
Especificación OpenAPI 3.1: https://gradusmusic.com/api-spec.yaml
Esquema JSON para el formato de entrada: https://gradusmusic.com/api/v1/notation/schema
Ejemplos de entrada canónicos: https://gradusmusic.com/api/v1/notation/examples
Documentación orientada a agentes: https://gradusmusic.com/llms-api.txt
git clone https://github.com/delmas41/gradusnotation
cd gradusnotation
npm install
npm run buildPara hacer una prueba de humo contra la API de producción:
node test-client.mjsAbre un issue en https://github.com/delmas41/gradusnotation/issues. Las contribuciones son bienvenidas: se prefieren PR pequeños y centrados.
MIT — Sean Johnson, Gradus School of Music Composition. Consulta LICENSE.
Search the Gradus music-theory knowledge base for authoritative source material. The corpus includes hand-authored curriculum prose, Bach chorale analysis (408 chorales), score commentaries on 50+ orchestral works, and primary historical sources from Fux (1725) through Boulanger.
WHEN TO USE: before generating notation if you need to look up a specific theory fact — typical voice leading for a Neapolitan-to-V resolution, idiomatic figured-bass realizations of a particular cadence, what makes a chromatic mediant feel like one composer's style versus another. Hitting this first prevents the agent from inventing chord progressions that are stylistically wrong.
WHEN NOT TO USE: for generic music vocabulary ("what is a chord?") that any LLM already knows; for non-theory queries like composer biographies, performance recommendations, or history dates — those are out of scope; for fetching actual score notation (use notation_render or notation_examples instead).
INPUT: provide EITHER topics (kebab-case tags) OR step (curriculum step 1-49). Topics are stronger; step is the fallback when you do not know the canonical topic tag. Both empty returns a MISSING_QUERY error.
OUTPUT (JSON): { ok: true, requestId, chunks: [{ id, sourceType, sourceId, title, content, composer?, era?, topics: string[], curriculumSteps: number[], tokenEstimate }], meta: { query, returnedCount, totalTokens, responseTimeMs }, attribution }. sourceType is one of: kg_concept, score_analysis, score_commentary, bach_chorale_analysis, composer, dictionary, curriculum, lesson_content, practicum, voice_leading, fugue, chorale_exercise, etc. Empty chunks: [] when nothing matched the topics — agent should fall back to its own knowledge or try a different topic tag.
EXAMPLE INPUT: { "topics": ["voice-leading", "deceptive-cadence"], "limit": 3 } TYPICAL LATENCY: 200-700 ms (one Voyage 3 embedding call + Supabase pgvector RPC).
| Name | Required | Description | Default |
|---|---|---|---|
| topics | No | Topic tags in kebab-case. Matched semantically via Voyage 3 Large embeddings plus a topic-overlap boost; exact-match is not required, so close synonyms work. Examples: ["voice-leading","deceptive-cadence"], ["chromatic-mediants"], ["sonata-form","second-theme"], ["figured-bass","6-4-2-chord"], ["fugue","stretto"], ["modulation","pivot-chord"]. | |
| step | No | Curriculum step number (1-49). Fallback when you do not know the topic tag. Maps to the Gradus 10-stage curriculum: Stage I 1-7 (single voice, intervals, scales), II 8-13 (counterpoint, all 5 species), III 14-16 (harmony, third voice), IV 17-18 (form, modulation), V 19-20 (fugue), VI 21-25 (classical style, sonata), VII 26-30 (Romantic harmony, augmented sixths), VIII 31-33 (Impressionist), IX 34-36 (20th century), X 37-40 (advanced). | |
| limit | No | Maximum chunks to return. Default 8 is right for most queries; raise for broad surveys, lower for tight context budgets. | |
| maxTokens | No | Token budget for the combined chunk content. Default 1500 fits comfortably in most agent context windows. The endpoint greedy-selects highest-similarity chunks within this budget. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Despite lacking annotations, the description thoroughly discloses behavioral traits: output JSON structure, sourceType enums, error on empty inputs, empty chunk behavior, and typical latency (200-700 ms). It covers all necessary operational aspects.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is moderately long but well-structured with labeled sections (WHEN TO USE, WHEN NOT TO USE, INPUT, OUTPUT, EXAMPLE INPUT, TYPICAL LATENCY). Every sentence adds value, and the purpose is front-loaded. No wasted words.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the absence of an output schema, the description fully explains the output format, including chunk objects with fields, sourceType list, empty chunks behavior, and attribution. It leaves no critical gaps for a search tool.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%, but the description adds significant context beyond the schema. It explains topics as kebab-case tags with semantic matching via Voyage, step as a fallback, and provides usage guidance for limit and maxTokens defaults (8 and 1500) with rationale.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states that the tool searches the Gradus music-theory knowledge base for authoritative source material, listing specific corpus contents. It also distinguishes itself from sibling tools by directly referencing notation_render and notation_examples for score notation, making its purpose unambiguous.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description explicitly provides 'WHEN TO USE' and 'WHEN NOT TO USE' sections, detailing scenarios such as looking up theory facts before generating notation, and excluding generic music vocabulary, non-theory queries, and score notation. It also suggests alternative tools for out-of-scope tasks.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Fetch canonical example inputs (single melody, two-voice counterpoint, chord progression, mixed rhythms with dynamics, string quartet snippet, tied notes across bar lines). Cache the result client-side; the response shape is stable.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are present, so the description carries full burden. It discloses that the response should be cached client-side and that the shape is stable, which is valuable behavioral context for an agent.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two concise sentences with front-loaded content. The first sentence lists examples clearly, and the second adds caching and stability info. No redundant text.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given no parameters or output schema, the description is sufficiently complete. It tells what the tool fetches and describes response characteristics, covering all necessary information for a simple fetch operation.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
With zero parameters, the baseline is 4. The description adds meaning by enumerating example categories, going beyond the empty schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool fetches canonical example inputs and lists specific examples like single melody and chord progression. It distinguishes from siblings such as knowledge_search, notation_render, notation_schema, and notation_validate by focusing on examples.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Usage is implied by listing examples, but the description lacks explicit guidance on when to use this tool versus other notation tools. No exclusions or alternatives are mentioned.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Render music notation from a JSON score. Returns inline SVG, MusicXML, and MIDI in one call. Use scientific pitches ("C4", "F#5", "Bb3") and duration codes (w h q 8 16 32 64 with optional dots). Bar lines are inferred from the time signature; notes that cross bar lines are split and tied automatically. Call notation_validate first if you are unsure your input is well-formed — validate is cheaper than render.
| Name | Required | Description | Default |
|---|---|---|---|
| title | No | Optional title rendered above the score. | |
| composer | No | ||
| tempo | No | ||
| timeSignature | No | ||
| keySignature | No | e.g. "C major", "G minor", "F# major". | C major |
| instruments | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description fully carries the burden of behavioral disclosure. It explains that bar lines are inferred from time signature and notes crossing bar lines are split and tied automatically. It also describes the pitch and duration format expected.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single paragraph that efficiently conveys purpose, output, input formats, behavior, and usage advice. It is front-loaded with the main action and each sentence adds value, though it could be slightly more concise.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's complexity and the lack of an output schema, the description provides good coverage of input formats and behavior. However, it does not explain all parameters (e.g., title, composer, tempo) in detail, leaving minor gaps.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is only 33%, but the description adds significant meaning: it explains scientific pitch notation ('C4', 'F#5'), duration codes (w, h, q, etc.), and the structure of notes (shortcut strings vs. objects). However, parameters like title, composer, and tempo are not elaborated beyond the schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose: 'Render music notation from a JSON score.' It specifies the output formats (SVG, MusicXML, MIDI) and distinguishes itself from sibling tools like notation_validate by advising to validate first.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description explicitly tells users when to use notation_validate instead ('if you are unsure your input is well-formed — validate is cheaper than render'). It also explains that bar lines are inferred and notes are automatically split, providing clear usage context.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Fetch the JSON Schema for the notation_render input shape. Cache the result client-side; this is stable across the v1 API.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations, so description carries full burden. Discloses stable API result and suggests client-side caching, adding value. No contradictions.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two sentences, no wasted words. Front-loaded with main action. Every sentence earns its place.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Adequate for a zero-parameter tool. Describes purpose and behavior. Could mention return format, but not essential given simplicity.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
No parameters, so baseline is 4. Description adds no parameter info, but none needed.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states it fetches the JSON Schema for notation_render input shape, specifying verb and resource. It distinguishes from siblings like notation_render (rendering) and notation_validate (validation).
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Implies usage context (fetch schema for notation_render) and advises caching due to stability. Does not explicitly exclude alternatives but given sibling tools, purpose is well-defined.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Pre-flight validate an input shape without rendering. Returns errors with concrete fix suggestions when input is malformed. Cheaper than notation_render — use this when iterating on input shape.
| Name | Required | Description | Default |
|---|---|---|---|
| title | No | ||
| composer | No | ||
| tempo | No | ||
| timeSignature | No | ||
| keySignature | No | ||
| instruments | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Without annotations, the description carries the burden of disclosing behavior. It mentions it returns errors with fix suggestions and is cheaper, but does not explicitly state that the tool is read-only, idempotent, or free of side effects—common expectations for a validation tool but not confirmed. More explicit behavioral context would be beneficial.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is two concise sentences. The first sentence states purpose and output; the second gives usage guidance. No repetition or filler. Essential information is front-loaded.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the absence of annotations and output schema, the description covers purpose and usage but omits detail on error types, fix suggestion format, input limitations, or edge cases. It provides a minimal but functional level of completeness, with room for more context.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema has 6 parameters with 0% description coverage; the description adds no parameter-specific meaning. While parameter names (title, composer, tempo, etc.) are self-explanatory, the description fails to clarify constraints, relationships, or how parameters influence validation. This is a significant gap.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description explicitly states the tool validates an input shape without rendering, distinguishing it from the sibling notation_render. It uses specific verbs ('validate') and identifies the resource ('input shape'), making the purpose unmistakable.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides clear guidance: 'Cheaper than notation_render — use this when iterating on input shape.' It tells the agent when to use (during iteration) and implies an alternative (notation_render for actual rendering).
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
knowledge_search4 fields changedInput schema / properties / limit / descriptionAdded value: +"Maximum chunks to return. Default 8 is right for most queries; raise for broad surveys, lower for tight context budgets."Input schema / properties / maxTokens / descriptionAdded value: +"Token budget for the combined chunk content. Default 1500 fits comfortably in most agent context windows. The endpoint greedy-selects highest-similarity chunks within this budget."Input schema / properties / step / descriptionPrevious value: -"Curriculum step number (1-49) as a fallback if you do not know the topic tag."New value: +"Curriculum step number (1-49). Fallback when you do not know the topic tag. Maps to the Gradus 10-stage curriculum: Stage I 1-7 (single voice, intervals, scales), II 8-13 (counterpoint, all 5 species), III 14-16 (harmony, third voice), IV 17-18 (form, modulation), V 19-20 (fugue), VI 21-25 (classical style, sonata), VII 26-30 (Romantic harmony, augmented sixths), VIII 31-33 (Impressionist), IX 34-36 (20th century), X 37-40 (advanced)."Input schema / properties / topics / descriptionPrevious value: -"Topic tags in kebab-case. Examples: [\"voice-leading\",\"deceptive-cadence\"], [\"chromatic-mediants\"], [\"sonata-form\",\"second-theme\"]."New value: +"Topic tags in kebab-case. Matched semantically via Voyage 3 Large embeddings plus a topic-overlap boost; exact-match is not required, so close synonyms work. Examples: [\"voice-leading\",\"deceptive-cadence\"], [\"chromatic-mediants\"], [\"sonata-form\",\"second-theme\"], [\"figured-bass\",\"6-4-2-chord\"], [\"fugue\",\"stretto\"], [\"modulation\",\"pivot-chord\"]."v0.1.1knowledge_searchnotation_examplesnotation_rendernotation_schemanotation_validateScored across 5 tools
Each tool has a clearly distinct purpose: knowledge_search for theory facts, notation_examples for example inputs, notation_render for rendering, notation_schema for schema retrieval, and notation_validate for input validation. There is no functional overlap.
Tools use a mix of noun_verb (knowledge_search, notation_render, notation_validate) and noun_noun (notation_examples, notation_schema) patterns. Additionally, one tool deviates from the 'notation_' prefix ('knowledge_search'), reducing consistency.
With 5 tools, the server is reasonably scoped for its purpose of music notation rendering and theory knowledge retrieval. It covers core functionality without being overly minimal or excessive.
The tool set covers search, retrieval of examples, input validation, schema access, and rendering. Minor potential gaps (e.g., no tool to list available examples or manage rendered outputs) are not critical for the stated domain.
Turn recordings, videos and sheet music into playable, editable scores. Transcribe audio or video (files or links) to sheet music, read sheet-music photos and PDFs, convert MIDI, MusicXML and ABC, edit notes in conversation with previews and undo, and export PDF, MIDI, MusicXML, guitar TAB, jianpu and audio. Hosted remote server with OAuth sign-in; one instrument, voice or solo piano is free.
Write lyrics in 100+ styles, score them, generate full songs with 4 engines, split stems. OAuth.
Render video and run AI media tasks from a single declarative JSON request.
Deterministic music theory for agents: analyze, voice, reharmonize, conduct — computed, not guessed
