Servidor MCP para herramientas de hora, zona horaria y duración.
Ejecútalo localmente a través de stdio (Claude Desktop, Claude Code, cualquier cliente MCP local), localmente a través de HTTP, o usa la instancia alojada — las mismas herramientas, el mismo código, tres formas de ejecutarlo.
Una instancia pública se ejecuta en Cloudflare Workers en https://time.somamcp.com/mcp:
claude mcp add --transport http temporal https://time.somamcp.com/mcpConstruido sobre somamcp, que proporciona la infraestructura MCP, la telemetría y los endpoints de salud/introspección para ambos runtimes. La lógica temporal es pura y funcional, usando functype.
Herramienta | Propósito |
| Hora actual como epoch, ISO-8601 UTC y hora de pared en cualquier zona IANA |
| Representa una marca de tiempo ISO-8601 en una zona horaria destino |
| Suma o resta una duración ISO-8601, con aritmética de meses consciente del calendario |
| Tiempo transcurrido entre dos marcas de tiempo, en unidades completas más un resumen legible |
somamcp también registra una herramienta info y los endpoints /health, /health/detail, /info y /dashboard.
La aritmética de meses satura en lugar de desbordar. add_duration sobre 2026-01-31 con P1M devuelve 2026-02-28, no 2026-03-03. Añadir «un mes» al final de un mes largo aterriza en el final del mes corto.
Los offsets se resuelven por instante, no por zona. America/New_York informa -04:00 en agosto y -05:00 en enero. El horario de verano (DST) proviene de la base de datos tz del runtime, por lo que aquí no hay una tabla de offsets que pueda quedar obsoleta.
Los errores llevan una pista. Una zona horaria desconocida devuelve el valor incorrecto y el formato esperado, para que un agente que llama pueda corregirse en lugar de volver a adivinar.
Related MCP server: mcp-datetimeday
Stdio es el modo predeterminado y el que esperan los clientes locales. Nada está alojado, nada escucha en un puerto — tu cliente lanza el proceso y se comunica con él a través de stdin/stdout.
claude mcp add temporal -- npx -y temporal-mcp-serverAñádelo a claude_desktop_config.json:
{
"mcpServers": {
"temporal": {
"command": "npx",
"args": ["-y", "temporal-mcp-server"]
}
}
}En macOS, ese archivo se encuentra en ~/Library/Application Support/Claude/claude_desktop_config.json; en Windows, en %APPDATA%\Claude\claude_desktop_config.json. Reinicia Claude Desktop después de editarlo.
Si prefieres no pasar por npm:
pnpm install
pnpm build
pnpm start # stdioLuego apunta tu cliente al punto de entrada compilado:
claude mcp add temporal -- node /absolute/path/to/temporal-mcp-server/dist/node.js{
"mcpServers": {
"temporal": {
"command": "node",
"args": ["/absolute/path/to/temporal-mcp-server/dist/node.js"]
}
}
}El paquete también instala un binario temporal-mcp-server, por lo que una instalación global (npm i -g temporal-mcp-server) te permite usar ese nombre directamente como comando.
El servidor habla JSON-RPC en stdout, así que puedes manejarlo manualmente:
printf '%s\n%s\n%s\n' \
'{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"smoke","version":"1.0.0"}}}' \
'{"jsonrpc":"2.0","method":"notifications/initialized"}' \
'{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"get_current_time","arguments":{"timezone":"Asia/Tokyo"}}}' \
| node dist/node.jsSolo el JSON-RPC va a stdout; los registros van a stderr, por lo que el pipe es seguro.
Para clientes que usan HTTP en streaming en lugar de stdio:
pnpm start:http # http://localhost:3333/mcp — override the port with PORTEs el mismo servidor y las mismas herramientas; solo difiere el transporte.
pnpm cf:dev # local workerd runtime
pnpm cf:deploy # build + edge-safety check + deploycf:deploy ejecuta pnpm build primero, que incluye check:worker — por lo que un bundle que contenga un built-in de Node falla antes de que nada llegue a Cloudflare.
Los despliegues se ejecutan a través de Cloudflare Workers Builds en lugar de GitHub Actions, por lo que no se almacena ningún token de API de Cloudflare en GitHub — Cloudflare se conecta al repositorio mediante su propia GitHub App.
Configúralo una vez en el panel (Workers & Pages → temporal-mcp-server → Settings → Build):
Campo | Valor |
Comando de despliegue |
|
Comando de compilación | (dejar vacío — |
Directorio raíz | (raíz del repositorio) |
Apuntar el comando de despliegue a un script del paquete mantiene la lógica de control en el control de versiones; el panel contiene una única línea estable. El nombre del Worker en el panel debe coincidir con name en wrangler.jsonc (temporal-mcp-server), o la compilación fallará.
La imagen de compilación incluye pnpm y respeta .nvmrc (la nuestra fija Node 24). Las ramas que no son de producción usan por defecto npx wrangler versions upload, por lo que los pushes de ramas producen versiones de vista previa sin tocar el despliegue en vivo.
El endpoint MCP está en /mcp. Para exigir un token de portador:
wrangler secret put MCP_AUTH_TOKENCon MCP_AUTH_TOKEN configurado, las llamadas no autenticadas a /mcp reciben un 401. Déjalo sin configurar y el endpoint será público — razonable para un reloj, no para mucho más.
vars opcionales: GIT_COMMIT y ENVIRONMENT se muestran mediante la herramienta info y /info.
La instancia pública se sirve desde un dominio personalizado:
claude mcp add --transport http temporal https://time.somamcp.com/mcpCon un token configurado, pásalo como encabezado:
claude mcp add --transport http temporal https://time.somamcp.com/mcp \
--header "Authorization: Bearer $MCP_AUTH_TOKEN"Comprobación de salud: https://time.somamcp.com/health.
pnpm cf:dev sirve lo mismo en http://localhost:8787/mcp, así que puedes apuntar un cliente a una instancia local de workerd antes de desplegar.
somamcp/edgeEl barrel raíz de somamcp reexporta helpers que importan node:fs. Importarlo desde un Worker arrastra los built-ins de Node al bundle. Por eso src/worker.ts importa somamcp/edge, y pnpm check:worker hace fallar la compilación si una importación node:, un built-in de Node sin prefijo o el especificador raíz somamcp llega al bundle del worker.
La comprobación recorre el grafo de importación real desde dist/worker.js en lugar de comparar nombres de archivo — el bundler eleva el código compartido con la entrada de Node a un chunk con un nombre generado, y un glob de nombres de archivo omitiría precisamente el archivo con más probabilidades de contener una fuga.
nodejs_compat está deliberadamente no habilitado en wrangler.jsonc. Si algún built-in de Node llegara, la compilación debería fallar ruidosamente en lugar de ser silenciosamente reemplazado.
alias en wrangler.jsoncxsschema (transitivo, a través de fastmcp) sondea cada librería de esquemas que soporta — valibot, effect, sury — mediante importación dinámica. Solo usamos zod, así que esas ramas nunca se ejecutan, pero esbuild todavía tiene que resolver los especificadores. Se les asigna un alias a un módulo vacío en lugar de instalar tres librerías no utilizadas.
src/
clock.ts pure time logic — Either<TemporalError, T>, no I/O, no globals
tools.ts MCP tool registration; takes a server, creates none
index.ts library surface (runtime-agnostic)
node.ts entry: somamcp -> stdio + httpStream
worker.ts entry: somamcp/edge -> export default { fetch }registerTemporalTools(server) recibe el servidor en lugar de construir uno, por lo que ambos puntos de entrada registran herramientas idénticas. Nada en clock.ts, tools.ts o index.ts toca process, el sistema de archivos ni ningún built-in de Node.
Los fallos son valores. Cada función que puede fallar en clock.ts devuelve Either<TemporalError, T>; la capa de herramientas convierte un Left en un resultado de error MCP. Nada depende del desenrollado de la pila, que es lo que permite que la misma lógica se ejecute sin cambios en ambos runtimes.
pnpm validate # format + lint + typecheck + test + build
pnpm test # 34 tests
pnpm check:worker # verify the worker bundle is edge-safetest/worker.spec.ts pasa objetos Request reales a través del manejador fetch del Worker mediante el protocolo de cable MCP, por lo que las roturas de integración aparecen en CI en lugar de después de un despliegue.
MIT
Add an ISO-8601 duration to a timestamp, in a timezone. Date units (Y/M/D) are calendar-aware: P1D across a DST boundary keeps the same wall-clock time the next day, and P1M on Jan 31 clamps to the end of February. Time units (H/M/S) are exact elapsed time, so PT24H really is 24 hours. Prefix the duration with '-' to subtract.
| Name | Required | Description | Default |
|---|---|---|---|
| duration | Yes | ISO-8601 duration, e.g. 'P1D', 'PT30M', 'P1Y2M3DT4H5M6S', or '-P1W'. | |
| timezone | No | IANA timezone governing the arithmetic — calendar units are applied to this zone's wall clock, and a timestamp with no 'Z' or offset is read as wall-clock time here. Defaults to UTC. | |
| isoTimestamp | Yes | ISO-8601 timestamp to shift. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description fully carries the behavioral disclosure burden. It explains DST boundary behavior, month-end clamping, exact elapsed time for time units, and the '-' subtraction prefix. This is exemplary transparency for a pure arithmetic tool.
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, front-loaded with the core action, then efficiently details edge-case behavior. Every sentence adds value; there is no filler or repetition of schema content.
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 moderately complex tool with no annotations and no output schema, the description covers the important behavioral nuances (DST, month-end, subtraction) thoroughly. It does not mention return value format or error cases, but those are relatively predictable for this operation, so the gap is minor.
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%, so the baseline is 3. The description adds meaningful semantics beyond the schema by explaining how duration units behave (calendar-aware vs exact elapsed time) and how timezone affects wall-clock arithmetic, which enriches the parameter meanings.
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: 'Add an ISO-8601 duration to a timestamp, in a timezone.' It clearly distinguishes this arithmetic tool from siblings like time_between (difference), convert_timezone (zone conversion), and get_current_time (current time).
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 context for when the tool is appropriate—duration arithmetic with calendar-aware behavior—but it does not explicitly name alternatives or state when not to use it. Sibling tool names imply the differentiation, but the description itself offers no direct guidance.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Render an ISO-8601 timestamp as wall-clock time in a target timezone.
| Name | Required | Description | Default |
|---|---|---|---|
| isoTimestamp | Yes | ISO-8601 timestamp to convert, e.g. '2026-08-17T14:30:00Z'. | |
| sourceTimezone | No | IANA timezone used to interpret `isoTimestamp` when it carries no 'Z' or offset. Ignored otherwise. Defaults to UTC. | |
| targetTimezone | Yes | IANA timezone identifier to render in. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It does not mention how source timezones are handled, DST behavior, return format, or error cases, leaving important operational details implicit.
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, concise sentence that front-loads the core action. Every word earns its place, with no redundancy or filler.
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 3-parameter tool with full schema coverage, the description is minimally adequate. However, without annotations or an output schema, it leaves the return format and the optional sourceTimezone edge cases unstated, which would help the agent use the tool more confidently.
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%, so the parameters are already well-documented. The description adds no additional parameter-specific meaning, so the baseline of 3 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 uses a specific verb ('Render') and clearly names both the input (ISO-8601 timestamp) and the desired outcome (wall-clock time in a target timezone). This distinguishes it from sibling tools like get_current_time and time_between.
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 no guidance on when to use this tool versus alternatives, and no exclusions or prerequisites. It simply describes the operation without contextualizing it against siblings such as add_duration or time_between.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Return the current time as an epoch, a UTC ISO-8601 timestamp, and the wall-clock time in a given timezone.
| Name | Required | Description | Default |
|---|---|---|---|
| timezone | No | IANA timezone identifier, e.g. 'America/New_York'. Defaults to UTC. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It states the tool returns time in three formats and accepts an optional timezone, but it does not explicitly state that the operation is read-only, non-destructive, or side-effect-free. For a simple read tool, this is adequate but not exhaustive.
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 sentence (18 words) that is front-loaded with the verb 'Return'. Every word contributes value, with no redundancy or filler.
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 simplicity (one optional parameter, no output schema), the description fully explains what the tool returns (epoch, ISO-8601 timestamp, wall-clock time in timezone). No additional information is necessary 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?
Schema description coverage is 100% (the only parameter 'timezone' already has a description including 'Defaults to UTC'). The tool description adds no new information beyond the schema, so the baseline score of 3 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 clearly states the verb 'Return' and the resource 'current time' with specific outputs (epoch, ISO-8601 timestamp, wall-clock time in timezone). It distinguishes itself from sibling tools like convert_timezone and time_between by focusing on the current moment rather than conversions or calculations.
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?
No guidance is given on when to use this tool versus alternatives (e.g., convert_timezone, time_between). There is no mention of context, prerequisites, or exclusions, leaving the agent to infer usage from the name alone.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Returns identity and build information for this server (name, version, build commit, runtime, capability counts)
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The annotations already declare readOnlyHint=true, so the description's disclosure of a read operation is consistent. It adds value by specifying the exact return content (name, version, build commit, runtime, capability counts), which is not captured in the annotations. No behavioral contradictions or hidden effects.
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 sentence that is front-loaded with the purpose ('Returns identity and build information for this server') and immediately lists the specific fields. Every word earns its place; there is no redundancy or fluff.
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 has no parameters, no output schema, and is read-only, the description fully covers what the agent needs to know: what it returns (identity and build info with listed fields). The context of sibling tools is clear, and no additional information is needed 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 tool has zero parameters and schema coverage is 100%, so the baseline is 4. The description does not need to explain parameters since there are none. It adds no additional parameter information but none is required.
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 returns 'identity and build information for this server' and lists the specific fields (name, version, build commit, runtime, capability counts). This clearly distinguishes it from the sibling tools, which are all time-related operations.
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 does not explicitly state when to use this tool versus alternatives. However, the sibling tools are all time-focused (get_current_time, convert_timezone, etc.), so the context implies this is the tool for server info. No explicit guidance on when not to use it is provided.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Exact elapsed time between two ISO-8601 timestamps, broken into days/hours/minutes/seconds plus a human-readable summary. All components share one sign, so they sum correctly; direction is 'past', 'future', or 'same'.
| Name | Required | Description | Default |
|---|---|---|---|
| to | Yes | End of the interval, as an ISO-8601 timestamp. | |
| from | Yes | Start of the interval, as an ISO-8601 timestamp. | |
| timezone | No | IANA timezone used to interpret timestamps that carry no 'Z' or offset. Defaults to UTC. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the burden. It discloses output breakdown (days/hours/minutes/seconds), a human-readable summary, consistent sign behavior, and a direction field. It does not cover edge cases like invalid timestamps or DST, but the disclosed details give good behavioral context.
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, both information-dense. The first delivers the core purpose, the second highlights the sign and direction semantics. No redundant or filler content.
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?
Even without an output schema, the description covers the main return components and behaviors. The tool is relatively simple, and the described details are sufficient for an agent to understand what will come back. Minor gaps like exact return type or error handling prevent a perfect score.
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%, so the schema fully documents all three parameters. The description adds no extra parameter-level meaning beyond restating the concept of the interval; 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 clearly states the tool computes 'exact elapsed time between two ISO-8601 timestamps', which is a specific verb-resource combination. It also distinguishes from siblings like add_duration or convert_timezone by focusing purely on time intervals.
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 purpose is self-evident: when an agent needs elapsed time between two timestamps, this is the tool. However, it does not explicitly call out when not to use it or mention alternatives, so it misses the bar for explicit guidance.
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.
v0.2.2add_duration1 field changedInput schema / properties / timezone / descriptionPrevious value: -"IANA timezone identifier, e.g. 'America/New_York'. Defaults to UTC."New value: +"IANA timezone governing the arithmetic — calendar units are applied to this zone's wall clock, and a timestamp with no 'Z' or offset is read as wall-clock time here. Defaults to UTC."convert_timezone1 field changedInput schema / properties / sourceTimezoneAdded value: +{
+ "description": "IANA timezone used to interpret `isoTimestamp` when it carries no 'Z' or offset. Ignored otherwise. Defaults to UTC.",
+ "type": "string"
+}time_between1 field changedInput schema / properties / timezoneAdded value: +{
+ "description": "IANA timezone used to interpret timestamps that carry no 'Z' or offset. Defaults to UTC.",
+ "type": "string"
+}v0.1.0add_durationconvert_timezoneget_current_timeinfotime_betweenScored across 5 tools
Each tool has a clearly distinct purpose: current time, timezone conversion, duration arithmetic, elapsed time, and server info. There is no meaningful overlap or ambiguity between them.
Most tools follow a consistent snake_case verb_noun pattern (add_duration, get_current_time, convert_timezone). The exception is 'info', which is a noun rather than a verb phrase, but this is a minor deviation.
Five tools is well-scoped for a focused temporal/time utility server. Each tool provides a distinct, non-redundant capability, and none feel unnecessary.
The core temporal workflows are covered: obtaining current time, converting timezones, adding durations, and measuring elapsed time. Missing advanced operations like listing timezones or extracting date components are minor gaps that agents can work around.
Timezone MCP — wraps timeapi.io (free, no auth)
timesheet.io MCP server - manage timers, projects, tasks and reports