Servidor MCP para la API v2 de Productive.io: proyectos, tareas, seguimiento de tiempo, planificación de recursos, finanzas, CRM e informes para una organización.
Productive expone unas 650 operaciones en 132 recursos. Convertirlas en 650 herramientas MCP saturaría la lista de herramientas de cualquier cliente, por lo que este servidor consta de doce herramientas impulsadas por un registro generado: las herramientas son genéricas y el registro sabe qué acepta realmente cada recurso.
Descubrimiento — productive_search_capabilities, productive_describe_resource, productive_check_connection, productive_describe_custom_fields
Lectura — productive_list (filtros, ordenaciones, inclusiones, paginación y los 26 endpoints de informes con agrupación), productive_get
Escritura — productive_create, productive_update, productive_delete, productive_run_action (150 verbos con nombre: archive, restore, approve, close, copy, finalize, send…), productive_track_time, productive_commit_operation
Autenticación por usuario (opt-in) — productive_connect, productive_status, productive_disconnect — ver Autenticación
Empieza con productive_search_capabilities. Los nombres de recursos de Productive son propios: un presupuesto es un deal, una columna de tablero es un workflow_status, una aprobación de parte de horas vive en time_entries — y adivinar cuesta llamadas.
Related MCP server: productive-mcp-rb2
src/productive/registry.generated.ts se deriva del documento OpenAPI publicado por Productive mediante scripts/generate-registry.mjs y se confirma en el repositorio, de modo que la CI nunca necesita la red y un cambio en la especificación aparece como un diff revisable. Para cada recurso registra los campos de filtro, las claves de ordenación, las claves de agrupación de informes, las relaciones incluibles, los atributos escribibles para crear y actualizar con los obligatorios marcados, y cada acción con nombre.
Eso es lo que permite que doce herramientas sigan siendo honestas. productive_describe_resource devuelve el contrato exacto de un recurso, y cada argumento se comprueba contra él antes de enviar una solicitud.
Regenéralo con npm run registry:generate (añade un argumento de ruta para usar una copia local de la especificación). El generador hace fallar la compilación si una clasificación escrita a mano — un nivel de riesgo, un indicador orientado al exterior, una operación bloqueada — ya no coincide con ninguna ruta de la especificación, de modo que un cambio de nombre upstream no puede eliminar silenciosamente una protección.
Todo lo que aparece aquí se verificó contra una organización real, porque la especificación y la API discrepan en lugares que importan.
El tiempo son minutos. El dinero son unidades menores. Una entrada de tiempo de 2 son dos minutos. Ambos vuelven también así.
Los filtros, ordenaciones e inclusiones desconocidos fallan de forma ruidosa. HTTP 400 con unsupported_filter, sort_param_unsupported, unsupported_include. Así que validarlos aquí es un error mejor, no una red de seguridad.
Los atributos de escritura desconocidos fallan en silencio. Un PATCH con un atributo mal escrito devuelve HTTP 200 y no cambia nada — indistinguible de un éxito. Por tanto, este servidor rechaza un atributo que el recurso no declara, en lugar de informar de una escritura que no ocurrió. Es lo más útil que hace el registro.
Hay exactamente seis operadores de filtro, en todos los campos: contains, eq, gt, lt, not_contain, not_eq. La especificación lista cuatro por campo y omite gt/lt, que sí funcionan; gte, lte, in, not_in, starts_with, ends_with, blank y present se rechazan todos con unsupported_filter_operation. No existe la comparación inclusiva, por lo que un rango inclusivo necesita los propios campos de filtro after/before o <field>_after/<field>_before del recurso.
page[size] tiene un tope de 200 y lo recorta en silencio. Pedir 500 devuelve 200 sin error. Los resultados incluyen total y nextPage para que una página no se confunda con la respuesta completa.
PATCH es genuinamente parcial. Los atributos omitidos conservan sus valores; no hace falta reenviar el registro completo.
data.type no se comprueba. Aplicar un PATCH a una tarea con type: "projects" tiene éxito y aplica el cambio. Este servidor envía el tipo correcto de todos modos.
Un 403 que dice que el id de organización "has to be provided" puede significar que era incorrecto, no que faltaba. El mismo código no_organization_id cubre una cabecera ausente y una organización a la que el token no puede acceder.
Una funcionalidad ausente responde 404, no 403. /boards devuelve 404 en una organización sin ella, lo que parece una ruta rota.
Los borrados pueden ser restaurables. Una tarea borrada aparece en deleted_items con item_type e item_id y puede restaurarse mediante la acción restore de ese recurso. Verificado solo para tareas: no asumas que se cumple para todos los tipos.
GET /users es el único endpoint con ámbito del llamante. Devuelve exactamente un registro — tú — y así es como este servidor identifica al propietario de un token. No existe /users/me; esa ruta devuelve 404. Cuidado con /organization_memberships: no está limitado a la organización fijada, sino que lista las membresías del llamante en todas las organizaciones a las que pertenece, así que su número de filas no es un recuento de personas.
No hay cabeceras de límite de tasa. Solo x-request-id, que los errores de este servidor citan. Reduce la velocidad ante un 429 en lugar de sondear el límite.
Cuatro interruptores, todos desactivados por defecto. Un servidor de solo lectura es el valor por defecto útil y seguro.
Interruptor | Cubre |
| Interruptor principal. Nada se modifica sin él. |
| Dinero, precios, nóminas, documentos que recibe un cliente: facturas, líneas de factura, pagos, recibos, gastos, órdenes de compra, propuestas, contratos, precios, tarifas, salarios, costes generales, tipos impositivos, cuentas bancarias, filiales. |
| Acceso y configuración de toda la organización: personas, membresías, conjuntos de permisos, equipos, invitaciones, campos personalizados, webhooks, integraciones, políticas de aprobación y de seguimiento de tiempo. |
| Borrados, además del control por niveles. |
Un único interruptor principal no basta para Productive: la misma API mueve una tarea, emite una factura y concede un conjunto de permisos, y esas son tres decisiones distintas. Un servidor en el que se confía para gestionar el trabajo de un proyecto no debería por ello poder enviar una factura.
PRODUCTIVE_ALLOWED_RESOURCES / PRODUCTIVE_DENIED_RESOURCES restringen aún más una instancia, y se aplican también a las lecturas: una instancia limitada al seguimiento de tiempo tampoco debería leer salarios.
Nunca expuestos en absoluto, independientemente de los interruptores: passwords, sessions, organization_subscriptions, los enlaces de compartir public/* sin autenticación, y PATCH /users/{id}/update_password. Estos están ausentes del registro en lugar de estar restringidos, de modo que ningún error de política puede reabrirlos.
El trabajo ordinario de un proyecto — una tarea, una entrada de tiempo, una reserva, un comentario — se escribe en una sola llamada. Exigir un apretón de manos en cada entrada de tiempo haría que el servidor fuera inutilizable para lo que la gente hace con más frecuencia.
Todo lo que tiene un radio de explosión mayor está preparado: la herramienta devuelve la solicitud exacta más un hash y no envía nada, y productive_commit_operation la ejecuta solo si la operación vuelve sin alteraciones. Eso cubre los niveles financiero y de administración, todos los borrados, cualquier cosa que salga de la organización y todas las acciones bulk_* — estas actúan sobre todos los registros que coinciden con un filtro, así que también se niegan a ejecutarse sin un filtro explícito.
Seis operaciones están marcadas como outward porque alcanzan a alguien fuera de la organización en el momento en que se ejecutan: invoices.send, invoices.send_einvoice, people.invite, people.resend, organizations.resend_code, y la creación de una invitation.
Dos modos. PRODUCTIVE_ORGANIZATION_ID es obligatorio en ambos, y nunca es un argumento de herramienta.
Cada persona vincula su propio token de Productive, de modo que Productive aplica sus permisos y registra su nombre en lo que hacen.
Esto importa más en Productive que en la mayoría de los sistemas. Productive atribuye el trabajo a personas: una entrada de tiempo pertenece a un person_id, y cada cambio queda sellado con el propietario del token en el registro de actividad — que es el registro con el que se defiende una factura de cliente. Con un token compartido, ese registro dice que la cuenta de servicio lo hizo todo.
PRODUCTIVE_PER_USER_AUTH=true
PRODUCTIVE_TRUST_FORWARDED_USER=true
PRODUCTIVE_ENCRYPTION_KEY=<min 16 chars>
PRODUCTIVE_STORE_PATH=/data/store.json
PRODUCTIVE_PUBLIC_BASE_URL=https://productive.example.com
# PRODUCTIVE_API_TOKEN deliberately unsetEl flujo:
El llamante ejecuta productive_connect y obtiene un enlace de un solo uso, válido durante 10 minutos, vinculado a su identidad.
Lo abre y pega un token que creó en Productive en Ajustes → Integraciones de API. El token va del navegador directamente al servidor, por lo que nunca entra en la transcripción de la conversación: un token de Productive equivale a un token de portador de toda su cuenta y, como se ha medido más abajo, suele llegar a más de una organización.
Antes de almacenarlo, el servidor llama a GET /users con ese token y el id de esta organización. Una llamada demuestra tres cosas: que el token es válido, que puede llegar a esta organización y de quién es. La página confirma entonces qué cuenta se vinculó.
Los tokens se cifran en reposo con AES-256-GCM, una fila por identidad verificada.
Bordes afilados:
La identidad solo proviene de la puerta de enlace. X-MCP-User se lee únicamente cuando PRODUCTIVE_TRUST_FORWARDED_USER=true, nunca de nada que controle el cliente MCP. Actívalo solo detrás de una puerta de enlace que establezca la cabecera a partir de un token validado y elimine una copia suministrada por el cliente — de lo contrario, un llamante puede nombrar cualquier identidad y actuar como ella.
Sin respaldo. Un llamante no registrado recibe NOT_CONNECTED, nunca el token compartido, incluso si PRODUCTIVE_API_TOKEN está definido. Un respaldo les daría derechos prestados, que es el fallo que este modo existe para eliminar.
/productive/enroll debe ser accesible desde el navegador del usuario, sin pasar por la puerta de enlace MCP: un navegador no puede llevar el token de portador de la puerta de enlace. Enruta /productive/* en PRODUCTIVE_PUBLIC_BASE_URL directamente al contenedor. Su seguridad es el token de estado de un solo uso y vinculado a la identidad.
Persiste PRODUCTIVE_STORE_PATH en un volumen y mantén estable PRODUCTIVE_ENCRYPTION_KEY: si la cambias, todos los tokens almacenados se vuelven indescifrables.
Si el correo electrónico del propio token de Productive difiere de la dirección de directorio del llamante, eso se notifica de forma visible en la página y en productive_status, y se conecta de todos modos. Establece PRODUCTIVE_REQUIRE_EMAIL_MATCH=true para rechazarlo en su lugar. Está desactivado por defecto porque quien pega el token de otra persona ya lo tiene, así que rechazarlo aporta poca seguridad, mientras que una cuenta de Productive con una dirección diferente es totalmente plausible.
La autenticación por usuario separa permisos y atribución, no organizaciones. La fijación de organización sigue aplicándose a todos.
Establece PRODUCTIVE_API_TOKEN a un token. Simple y adecuado para stdio o un único operador — pero entonces cada llamante actúa como el propietario de ese token, con sus permisos, y el registro de actividad de Productive acredita cada cambio a esa persona.
PRODUCTIVE_TRUST_FORWARDED_USER sigue ayudando aquí: las escrituras relativas a una persona (una entrada de tiempo, una reserva)
se asignan por defecto al llamante resuelto en lugar del propietario del token, y el servidor se niega a adivinar cuando
la dirección no coincide con nadie o con más de una persona. productive_check_connection nombra al propietario del token
en cualquier caso, por lo que la atribución nunca es una sorpresa.
Una instancia sirve exactamente a una organización. X-Organization-Id proviene del entorno y
nunca es un argumento de herramienta, por lo que ninguna ruta de código — incluidas las herramientas genéricas — puede alcanzar a otro inquilino.
Ejecuta una segunda instancia para una segunda organización; la imagen es la misma.
Esto no es teórico. Un solo token llega habitualmente a varias organizaciones: en la cuenta contra la que
se desarrolló, GET /organizations devolvió tres, y cambiando solo la cabecera se movía
entre ellas (las otras dos respondieron 403 subscription_expired, no "no encontrado"). La cabecera es
todo el límite, por eso está fijada en lugar de pasarse — y por eso un token por usuario registrado
se verifica contra esta organización antes de almacenarse.
Consulta .env.example. Las dos variables obligatorias son PRODUCTIVE_API_TOKEN (Ajustes → Integraciones
de API en Productive; hereda los permisos del usuario que la crea) y
PRODUCTIVE_ORGANIZATION_ID (el id numérico en tu URL de Productive).
Establece PRODUCTIVE_AUDIT_LOG para añadir una línea JSON por cada intento de mutación, incluidos los que la política
rechazó. Los cuerpos de las solicitudes no se registran deliberadamente: contienen salarios, tarifas y datos
personales, y una pista de auditoría que debe protegerse tan estrechamente como el sistema de origen tiende a no
leerse.
npm install
npm run dev # stdio
npm run dev:http # streamable HTTP on :3000/mcp (stateless), /healthz open
npm test
npm run smoke:live # reads a real organization; stages one write, commits nothingImágenes de Docker: ghcr.io/borgels/mcp-server-productive (publicadas al hacer push a main).
Apache-2.0.
Verify the token and organization and report what this deployment may do. Also names the person the token belongs to — every change Productive records is credited to them, whoever asked — and the caller identity, when a gateway forwards one.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The readOnlyHint annotation already signals safety, and the description adds relevant behavioral context: token identity is attributed to every recorded change, and caller identity may be forwarded by a gateway. This goes beyond the annotation without contradicting it.
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 sentences with the core action first, followed only by consequential detail about attribution and caller identity. There is no fluff or repetition of the title.
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?
For a parameterless tool, the description conveys the main outputs: deployment capabilities, token owner, and caller identity. It does not specify exact response formatting or failure behavior, but the tool's simplicity and readOnly annotation make this a minor gap.
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 tool has zero parameters and the empty schema fully describes the input surface, so there is nothing for the description to add about parameter semantics. The baseline of 4 applies.
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 opens with a specific verb and resource: 'Verify the token and organization' and states what the tool reports (deployment capabilities, token owner, caller identity). This makes its role distinct from sibling data-operation tools like productive_get or productive_create.
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 clearly establishes the context — validating the Productive token and organization and surfacing what the deployment can do — so an agent knows to call it for connection/auth checks. It does not explicitly name alternatives or when-not conditions, but the absence of parameters and the verification purpose make the intended use obvious.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Execute a write that a prepare step staged. Pass the operation object back exactly as it was returned: it is hashed, and an altered operation is refused rather than run. Show the change to the person who asked for it before committing.
| Name | Required | Description | Default |
|---|---|---|---|
| operation | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Beyond the minimal readOnlyHint=false annotation, the description discloses two important behaviors: the operation is hashed and altered operations are refused, and the change must be shown to the requester before committing. This is valuable safety-relevant 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?
Three short, purposeful sentences with no fluff. Purpose, usage constraint, and mandatory approval step are all front-loaded and directly actionable.
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?
The description covers what matters most for a commit operation: exact object reuse, hash enforcement, and user approval. It does not describe the success response or explicitly name a prepare tool, but the workflow is sufficiently constrained for correct invocation.
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 schema has 0% description coverage, and the single nested 'operation' parameter has no field-level documentation. However, the description compensates by instructing the agent to pass the exact object back and explaining the hash invariant, which materially reduces the risk of misuse despite lacking field-by-field semantics.
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?
Description states a specific verb and resource: 'Execute a write that a prepare step staged.' This clearly identifies the tool as the commit step for a staged operation, distinguishing it from direct create/update/delete siblings by the staged-operation framing.
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?
Gives clear usage context: use after a prepare step, pass the operation object back exactly as returned, and show the change before committing. It does not explicitly name alternatives or say when not to use it, but the staged-operation workflow is unmistakable.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Create a record on any writable resource. Required attributes are checked first, and so is every attribute name. On a financial or administrative resource this returns a staged operation to review rather than creating anything — pass it to productive_commit_operation to go through with it.
| Name | Required | Description | Default |
|---|---|---|---|
| resource | Yes | ||
| attributes | No | Attributes to write, by their Productive names. Validated against the resource contract first, because Productive answers 200 and ignores an unknown attribute rather than reporting it. Custom-field values go in a nested `custom_fields` object keyed by field id. | |
| relationships | No | JSON:API relationships, where the resource accepts them. Most links are set through an `_id` attribute instead. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations only mark readOnlyHint=false, so the description carries the burden of explaining mutation nuance. It discloses the staging behavior, the validation-before-write behavior, and the fact that commit is required on certain resources. This is meaningful behavioral context beyond the schema.
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?
Three sentences, no filler, with the core action first, then validation, then the exception case. Every sentence adds actionable information.
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?
For a create tool with no output schema, it explains the essential invocation path, including when commit is required. It could say more about return shape or failure behavior, but the description and schema together cover the necessary details.
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 schema already describes attributes and relationships in detail, covering 2 of 3 parameters. The tool-level description adds no parameter-specific detail; `resource` remains minimally described, though it is self-evident and explained at a high level as 'any writable resource.'
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?
States exactly what it does: create a record on any writable resource. The explicit mention of productive_commit_operation distinguishes it from update/delete/list/get siblings. No ambiguity about the operation performed.
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?
Explains the resource-type condition that changes behavior: financial/administrative resources stage an operation rather than creating, and directs the agent to productive_commit_operation to finalize. It doesn't enumerate exclusions versus update/delete, but for a create tool this is sufficient guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Delete a record. Needs its own switch beyond ordinary writes, and always returns a staged operation first. Many Productive types are meant to be archived rather than deleted — check productive_describe_resource for an archive action before reaching for this.
| Name | Required | Description | Default |
|---|---|---|---|
| id | Yes | ||
| resource | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare destructiveHint=true and readOnlyHint=false, so the description need not restate that. It adds non-obvious behavioral facts: deletion requires a separate switch and always returns a staged operation first. This goes beyond annotations and meaningfully sets expectations.
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 three concise sentences with the core purpose first, followed by two high-value caveats. Every sentence earns its place; there is no filler or repetition of schema/annotations.
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?
For a destructive tool with two simple parameters and no output schema, the description provides essential behavior: staged operation first, archive exception, and switch requirement. The main gap is parameter semantics, but the operational guidance is strong enough to make the tool safely invocable.
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 description coverage is 0%, and the description does not define the `resource` or `id` parameters. 'Delete a record' hints that resource is the record type and id is its identifier, but this is too vague to clarify valid resource types or how the id is used. The description adds minimal value beyond the parameter names.
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 states a specific action ('Delete') and resource ('a record'), making the tool's purpose clear. It also distinguishes the tool from create/update operations and explicitly references the archive alternative, differentiating it from a sibling behavior.
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?
It explicitly instructs the agent to check productive_describe_resource for an `archive` action before using delete, giving clear when-to-use-vs-alternative guidance. It also notes the requirement for a special switch beyond ordinary writes, which is a practical precondition for invocation.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List the organization's custom fields with the id each value is written under, the object it applies to, and its options. Custom-field values are keyed by numeric id, not by display name — a name used as a key is accepted with 200 and stored nowhere.
| Name | Required | Description | Default |
|---|---|---|---|
| appliesTo | No | Narrow to fields on one object, matched against Productive's own wording for it. That wording is not this server's resource key — fields on people report "employees" — so call this without an argument first and pick from appliesToValues. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Beyond the readOnlyHint annotation, the description discloses a critical behavioral quirk: custom-field values are keyed by numeric id, not display name, and using a name as a key is silently accepted with 200 but stored nowhere. This is exactly the kind of non-obvious behavior that prevents an agent from making a subtle mistake.
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 sentences with no filler. It front-loads the purpose and return fields, then adds the essential behavior warning. 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?
For a read-only listing tool with one optional parameter, the description is complete: it names the resource, lists the output components, and flags the keying gotcha. The absence of an output schema is mitigated by the explicit statement of what is returned, and the readOnlyHint annotation covers safety expectations.
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 100% description coverage and already explains appliesTo in detail, including the 'employees' example and the instruction to call without an argument first. The tool description itself adds no additional parameter semantics beyond the schema, so the baseline of 3 is appropriate.
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 uses a specific verb ('List'), names the resource ('organization's custom fields'), and specifies what is returned: id, object, and options. It clearly distinguishes this from siblings like productive_describe_resource, which would describe a resource's schema rather than enumerate custom fields.
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 implies its usage context clearly: call it to list custom fields and understand how values are keyed. The input schema adds explicit guidance to call without an argument first and then pick from appliesToValues, which is practical usage direction. It does not explicitly name alternative tools or state when not to use it, so it stops short of a 5.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
The exact contract for one resource, derived from Productive's own OpenAPI document: every filter field, the operators any field accepts, sort keys, report group keys, includable relationships, the writable attributes for create and update with the required ones marked, and any named actions. Read this before writing — it is the difference between a write that lands and a 200 that changed nothing.
| Name | Required | Description | Default |
|---|---|---|---|
| resource | Yes | Resource key, e.g. "tasks", "time_entries", "reports/time_reports". |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint=true, so the safety profile is covered. The description adds valuable behavioral context: output is derived from Productive's own OpenAPI document and includes specific contract elements, plus a warning about write no-ops that goes beyond the annotation.
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 dense but efficient: one sentence enumerates the full contract contents, and a second short sentence delivers a high-impact usage warning. Every phrase contributes meaning, with no filler or repetition of the schema.
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?
For a simple one-parameter read-only introspection tool, the description fully specifies what output the agent should expect and when to call it. The lack of an output schema is compensated by the detailed enumeration of returned contract components.
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 description coverage is 100%, and the schema already documents the single 'resource' parameter with examples. The description adds no additional syntax, validation, or format details for the parameter, so it stays at the baseline where the schema does the heavy lifting.
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 states a precise verb-object relationship: it provides 'the exact contract for one resource' and enumerates the concrete contents such as filter fields, operators, sort keys, and writable attributes. This clearly distinguishes it from siblings like productive_get, productive_list, and productive_describe_custom_fields.
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 the agent to 'Read this before writing' and contrasts a successful write with 'a 200 that changed nothing,' giving a clear trigger for use. It does not explicitly name alternative tools or when not to use it, but the intended context is unambiguous.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Read one record by id, optionally pulling related records in the same call rather than following relationship stubs one at a time.
| Name | Required | Description | Default |
|---|---|---|---|
| id | Yes | ||
| raw | No | ||
| include | No | ||
| resource | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The readOnlyHint annotation already establishes safety, and the description adds meaningful behavior: optional eager-loading of related records in the same call. It does not explain raw mode or default include behavior, but the core read semantics are transparent and align with annotations.
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, well-structured sentence that front-loads the core purpose and then adds the key optional behavior. There is no redundant or filler language.
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?
With no output schema and no parameter descriptions, the description is too sparse to fully support correct invocation. The agent still has to guess what 'resource' values are valid, what 'raw' does, and what the response shape looks like. The relationship-include guidance is helpful but not enough.
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 description coverage is 0%, so the description must compensate for all four parameters. It explains the purpose of 'id' and loosely hints at 'include', but it never clarifies 'resource' or 'raw', which are both undocumented in the schema and the description. 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 uses a specific verb ('Read') and clear object ('one record by id'), which distinguishes it from sibling tools like productive_list and productive_create. The added phrase about pulling related records also separates it from naive relationship-stub traversal.
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 conveys when to use the tool: when you need a single record by id and optionally want related records in one call rather than following stubs. It does not explicitly exclude alternatives like productive_list for multiple records, so it stops short of full when-to-use guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List any Productive resource with server-side filtering, sorting, includes and paging. Also how the 26 report endpoints are read: use resource "reports/time_reports" (and the rest) with group to get aggregated rows instead of records. Results are flattened out of JSON:API — attributes hoisted, included records inlined into their relationship — with total and nextPage so a page is never mistaken for the whole answer.
| Name | Required | Description | Default |
|---|---|---|---|
| raw | No | Return Productive's JSON:API envelope untouched instead of flat records. | |
| page | No | ||
| sort | No | Sort keys; prefix with "-" for descending, e.g. ["-created_at"]. | |
| group | No | Report aggregation key. Only the reports/* resources accept one. | |
| filters | No | Server-side filters. A bare value means equality: {"assignee_id":"5"}. An array is comma-joined: {"id":["1","2"]}. An operator object narrows further: {"created_at":{"gt":"2026-01-01"}} — the ONLY operators Productive accepts are contains, eq, gt, lt, not_contain, not_eq. There is no gte/lte, so for an inclusive date range use the resource's own after/before filter fields instead. Logical groups nest: {"$op":"and","0":{"assignee_id":{"eq":"5"}},"1":{"company_id":{"eq":"7"}}}. Field names are validated against the resource — call productive_describe_resource for the list. | |
| include | No | Relationships to pull in the same call, e.g. ["assignee","project"]. | |
| resource | Yes | One of 132 resource keys; see productive_search_capabilities. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The description adds significant behavioral detail beyond the readOnlyHint annotation: results are flattened out of JSON:API, attributes are hoisted, included records are inlined, and responses include total and nextPage so the agent understands pagination is partial. It also clarifies that report resources return aggregated rows rather than records. This goes well beyond the minimal safety signal.
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 concise at three sentences and front-loads the core purpose. Each sentence provides meaningful information, though the phrase 'Also how the 26 report endpoints are read' is slightly awkward and could be streamlined. Overall it is efficient without unnecessary detail.
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?
With no output schema, the description explains the return shape well: flattened records, inlined relationships, total, and nextPage. It also covers the special report behavior and aggregated rows. For a complex 7-parameter listing tool with 132 possible resources, this is sufficiently complete to guide correct invocation.
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 description coverage is 86%, so the input schema already documents most parameters well. The description adds a little context around group for report endpoints, but it mostly summarizes capabilities like filtering and paging rather than meaningfully extending the schema's parameter documentation. Baseline 3 is appropriate because the schema carries the parameter-semantic burden.
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 lists any Productive resource with filtering, sorting, includes, and paging. It also explicitly handles the special behavior of the 26 report endpoints, distinguishing this from siblings like productive_get and productive_search_capabilities. The verb 'List' and resource scope are 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 gives clear guidance on when to use this tool, including the specific instruction to use resource 'reports/time_reports' with group for aggregated report rows. It provides the main usage context but does not explicitly enumerate exclusions or alternatives such as 'use productive_get for a single record.' The context is clear enough that an agent can choose correctly in most cases.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Run one of Productive's named verbs: archive, restore, close, open, approve, reject, copy, finalize, send, reposition and the rest. Actions named bulk_* act on every record the filter matches rather than one, and refuse to run without an explicit filter. productive_describe_resource lists the actions a resource has.
| Name | Required | Description | Default |
|---|---|---|---|
| id | No | Required for actions that act on one record. | |
| action | Yes | Action name, e.g. "archive", "approve", "send". | |
| filters | No | Server-side filters. A bare value means equality: {"assignee_id":"5"}. An array is comma-joined: {"id":["1","2"]}. An operator object narrows further: {"created_at":{"gt":"2026-01-01"}} — the ONLY operators Productive accepts are contains, eq, gt, lt, not_contain, not_eq. There is no gte/lte, so for an inclusive date range use the resource's own after/before filter fields instead. Logical groups nest: {"$op":"and","0":{"assignee_id":{"eq":"5"}},"1":{"company_id":{"eq":"7"}}}. Field names are validated against the resource — call productive_describe_resource for the list. | |
| resource | Yes | ||
| attributes | No | Attributes to write, by their Productive names. Validated against the resource contract first, because Productive answers 200 and ignores an unknown attribute rather than reporting it. Custom-field values go in a nested `custom_fields` object keyed by field id. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations only provide readOnlyHint=false, so the description carries the behavioral burden. It discloses two non-obvious behaviors: bulk_* actions operate on every record the filter matches, and they refuse to run without an explicit filter. The schema also surfaces that Productive answers 200 and silently ignores unknown attributes, adding important behavioral context beyond the annotation.
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 three sentences with no filler. It front-loads the core purpose, then immediately covers the most dangerous edge cases (bulk scope and filter requirement), and ends with a useful discovery pointer. 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?
For a complex 5-parameter tool with nested filter objects and no output schema, the description and schema together cover the essential invocation details: valid action classes, bulk semantics, filter requirements, field validation, and discovery via productive_describe_resource. The only notable gap is the lack of any mention of what the response looks like or error behavior beyond the silent-attribute-ignore note, which is a minor omission for an action-triggering 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 description coverage is roughly 80%, and the schema already provides rich semantics for filters, attributes, and id. The description adds the bulk-vs-single distinction and the explicit-filter requirement, which complements the filters parameter. Given this high schema coverage, the baseline of 3 is appropriate; the description does not need to re-document each parameter.
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 names a specific verb ('Run') and a clear resource class ('one of Productive's named verbs'), then lists concrete examples: archive, restore, close, open, approve, reject, copy, finalize, send, reposition. It also distinguishes single-record actions from bulk_* actions, so the tool's purpose is unmistakable and well differentiated from the sibling CRUD tools.
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 clearly states that bulk_* actions require an explicit filter and calls out productive_describe_resource as the way to discover which actions a resource supports. It does not, however, give explicit when-not-to-use guidance against siblings like productive_update or productive_delete, so it falls short of a 5.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Find the right resource, tool or guide for a subject. Start here: Productive names its resources its own way (a budget is a deal, a person is a person, a board column is a workflow_status), and guessing wastes calls. Returns matching tools, workflow guides, and resources with the operations each supports.
| Name | Required | Description | Default |
|---|---|---|---|
| limit | No | ||
| query | Yes | Subject to search for, e.g. "invoice", "capacity", "custom field", "time". |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With readOnlyHint=true already in annotations, the description adds useful behavioral context: it returns matching tools, workflow guides, and resources along with the operations they support. It also discloses the important quirk that resource names may not match user expectations. No contradiction with annotations.
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?
Three sentences, front-loaded with the core purpose, and every sentence adds value: what it finds, where to start, and what it returns. The naming examples are illustrative without being verbose.
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?
For a read-only search tool with only two parameters and no output schema, the description covers purpose, scoping context, and return categories well. It does not detail result shape or empty-result behavior, but those are not critical for selecting and invoking this tool correctly.
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 description reinforces that 'query' is a subject to search for, but the schema already covers that with examples. The 'limit' parameter is not described in the description, and with only 50% schema description coverage, the description does not fully compensate. The parameter is simple enough that the gap is moderate rather than severe.
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 states a specific verb ('Find') and resource ('resource, tool or guide for a subject') and clearly distinguishes this as the starting point for discovery. It explains the tool's role relative to siblings by highlighting Productive's naming conventions, so an agent knows this is a search/capability lookup, not a data operation.
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?
'Start here' explicitly tells the agent when to use this tool, and the mention of Productive's non-obvious naming ('a budget is a deal') explains why guessing would waste calls. It does not name alternative sibling tools or state when not to use it, but the contextual guidance is clear enough.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Log a time entry without looking up what Productive insists on. Time is stored in MINUTES — pass hours and it converts. The service (the budget line the work bills to, not the task) is resolved from the task when not given, and the person is resolved from the authenticated caller, so time lands on the right person even though the server holds one shared token.
| Name | Required | Description | Default |
|---|---|---|---|
| date | Yes | Date of the work, YYYY-MM-DD. | |
| note | No | ||
| hours | No | Time worked in hours; converted to minutes. | |
| taskId | No | Task worked on. Also used to resolve the service. | |
| minutes | No | Time worked, in minutes. | |
| personId | No | Whose time this is. Defaults to the resolved caller. | |
| serviceId | No | Service (budget line) to bill to; overrides the task lookup. | |
| billableMinutes | No | Billable portion, if it differs from the time worked. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With only readOnlyHint=false in annotations, the description carries the full burden of behavioral disclosure. It clearly reveals that hours are converted to minutes, the service is derived from the task when absent, and the person is resolved from the authenticated caller despite a shared server token. This is specific and valuable transparency.
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?
Three sentences, each earning its place: the first states the action, the second explains the critical unit conversion, and the third clarifies lookup defaults and authentication behavior. It is dense without being bloated, and the essential action 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?
For an 8-parameter tool with no output schema, the description covers the core call flow well: what to log, how units are handled, which IDs can be omitted, and how resolution works. It does not describe the return value or what happens when conflicting inputs like hours and minutes are both supplied, but these are minor gaps given how much is already explained.
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 88%, so the schema already documents most parameters. The description adds real value by explaining the hours-to-minutes conversion, the distinction between service and task, and the fallback resolution behavior for serviceId and personId. It does not explain every edge case, such as providing both hours and minutes, but it meaningfully supplements 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 opens with a specific verb and resource: 'Log a time entry.' It goes beyond the title by explaining the key behavior that makes this tool distinct from generic create/list tools: it avoids manual lookups and handles conversions, making the intent 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 gives clear practical context: service can be omitted and is resolved from the task, person can be omitted and is resolved from the caller. This tells an agent when it can avoid passing certain parameters, though it does not explicitly name alternatives or state when not to use the tool.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Update a record. PATCH semantics, verified live: attributes you omit keep their values, so there is no need to resend the whole record. On a financial or administrative resource this returns a staged operation instead of writing.
| Name | Required | Description | Default |
|---|---|---|---|
| id | Yes | ||
| resource | Yes | ||
| attributes | No | Attributes to write, by their Productive names. Validated against the resource contract first, because Productive answers 200 and ignores an unknown attribute rather than reporting it. Custom-field values go in a nested `custom_fields` object keyed by field id. | |
| relationships | No |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With only readOnlyHint=false provided by annotations, the description carries a meaningful behavioral burden and does well: it discloses PATCH semantics, confirms omitted fields are retained, and warns that certain resources return a staged operation instead of writing. It does not cover permissions or reversibility, but it adds real behavioral facts beyond the annotations.
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 short and front-loaded with the key PATCH behavior. The first sentence duplicates the title, which costs a little, but every other sentence carries useful operational meaning without padding.
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?
The description covers core PATCH semantics and the staged-operation caveat, but it does not explain what a staged operation should be followed by, especially given the sibling productive_commit_operation. There is also no output schema and the description does not say what a normal update returns, and relationships remain undocumented. This leaves the agent needing to infer important follow-up behavior.
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 low at 25%, and the description does add value for the attributes parameter by explaining that omitted attributes keep their values. However, it gives no semantic guidance for resource, id, or relationships, and the custom_fields nesting behavior lives only in the schema, not the description.
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 states a clear action on a record and immediately clarifies PATCH semantics, which distinguishes it from read-only siblings like productive_get and productive_list. It is slightly repetitive with the title but unambiguous about the operation's purpose.
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 gives useful context: partial updates preserve omitted attributes, so no full resend is needed, and financial/admin resources return staged operations instead of writing. However, it does not explicitly say when to choose this over productive_create, productive_delete, or productive_run_action, so usage versus alternatives is mostly implied.
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.
v0.2.0productive_check_connectionproductive_commit_operationproductive_createproductive_deleteproductive_describe_custom_fieldsproductive_describe_resourceproductive_getproductive_listproductive_run_actionproductive_search_capabilitiesproductive_track_timeproductive_updateScored across 12 tools
Each tool targets a distinct operation: reading one record, listing, creating, updating, deleting, running a named action, staging/committing writes, tracking time, or performing a meta/introspection task. Even the closer pairs like get/list and describe_resource/describe_custom_fields are clearly separated by resource scope and purpose.
All tools follow the snake_case productive_<verb> pattern, with most using verb_noun (productive_check_connection, productive_run_action) and a few using bare verbs (productive_get, productive_list). This is consistent and predictable, with no style mixing or vague duplicate verbs.
Twelve tools is well within the ideal scope for a broad API integration server. The set covers generic CRUD, named actions, time tracking, staged writes, and meta/discovery tools without ballooning into redundant or overly granular endpoints.
The server provides full lifecycle coverage: read, list, create, update, delete, run actions, and commit staged operations, plus support for custom fields and resource contracts. The meta tools fill the gaps that typically cause agent failures, such as knowing resource-specific filters, writable attributes, and custom-field keying.
Agent-complete, permission-scoped product operations for Priorify workspaces.
Supervised API-write gateway for AI agents with policy, human approval and execution receipts.
Read time entries, projects, clients, tasks and invoices; log and update tracked time.
Query and analyze your pylo data. Changes become proposals you approve in the pylo admin UI or API.