Skip to main content
Glama
deenesjakoruzh
theYahia

claude-webcache

by theYahia
OverviewSchemaSourceRelated ServersDiscussionsAdmin
JavaScript
Local

claude-webcache

npm license downloads

Caché WebFetch entre sesiones para Claude Code.

El WebFetch integrado de Claude Code almacena los resultados en caché durante 15 minutos dentro de una misma sesión. claude-webcache extiende esto entre sesiones, indefinidamente (TTL de 7 días, configurable).

Open new session  ->  your past fetches are still there.
Cache hit         ->  instant.
Cache miss        ->  same as built-in WebFetch.

Por qué

Cada vez que vuelves a obtener las mismas URLs entre sesiones (documentos, referencias de API, páginas de investigación), pagas el coste total de la obtención cada vez. La caché de 15 minutos dentro de la sesión se elimina antes de tu próximo sprint. claude-webcache mantiene las obtenciones para que la segunda sesión acceda a la caché en su lugar.

Flujo CACHE_MISS: WebFetch + cache_store en la primera sesión Flujo CACHE_HIT: acierto instantáneo, sin WebFetch en la segunda sesión

Related MCP server: ClaudeX

Instalación

Opción 1 -- Plugin de Claude Code (recomendado)

Una línea, en tu terminal:

claude plugin marketplace add theYahia/claude-webcache && claude plugin install claude-webcache@theyahia

Luego añade el patrón de uso a tu ~/.claude/CLAUDE.md (consulta Patrón de uso).

💡 ¿Por qué subcomandos de CLI y no el /plugin install de la TUI? El flujo de la TUI actualmente se dirige a través del backend remoteMarketplaceClient de Claude Code, que tiene un error abierto en el lado del servidor que rechaza todas las fuentes de plugins de terceros con Failed to install: This plugin uses a source type your Claude Code version does not support. Consulta anthropics/claude-code#41653 y unas 20 incidencias abiertas relacionadas. Los subcomandos de CLI anteriores utilizan NativeMarketplaceReader localmente y no se ven afectados.

**Alternativa (si incluso los subcomandos de CLI fallan en tu configuración):

git clone https://github.com/theYahia/claude-webcache && claude --plugin-dir ./claude-webcache/plugin

Esto evita el pipeline del marketplace por completo.

Opción 2 -- npm global

npm i -g @theyahia/claude-webcache

Requiere Node.js 22.5+ (utiliza node:sqlite integrado, sin dependencias nativas).

Luego regístralo manualmente en ~/.claude/settings.json (reemplaza la ruta con el resultado de npm root -g):

{
  "mcpServers": {
    "claude-webcache": {
      "command": "node",
      "args": ["/path/from/npm-root-g/claude-webcache/scripts/mcp-server.cjs"]
    }
  },
  "hooks": {
    "SessionStart": [
      {
        "matcher": "startup|clear|compact",
        "hooks": [
          { "type": "command", "command": "node /path/from/npm-root-g/claude-webcache/scripts/hook-stats.cjs" }
        ]
      }
    ]
  }
}

Opción 3 -- clonar (colaboradores)

Consulta CONTRIBUTING.md.

Patrón de uso

Añádelo a tu ~/.claude/CLAUDE.md para que Claude dirija las obtenciones web a través de la caché:

## WebFetch caching (claude-webcache)

Before calling WebFetch, call `cached_fetch(url, prompt)` first.
- If it returns text -> use that, do NOT call WebFetch.
- If it returns `[CACHE_MISS] <url>` -> call WebFetch as normal, then call `cache_store(url, prompt, output)` with the result.

Eso es todo. Misma URL + mismo prompt en cualquier sesión futura = acierto de caché instantáneo.

Herramientas (MCP)

Herramienta

Argumentos

Devuelve

cached_fetch

url, prompt

texto en caché, o [CACHE_MISS] <url>

cache_store

url, prompt, output

stored

cache_stats

--

{ total, hits, last }

cache_list

limit?

URLs recientes (las más recientes primero)

Almacenamiento

SQLite en ~/.webcache/cache.db (modo WAL, seguro para concurrencia). Clave de caché = SHA256(url + "|" + prompt).

Campo

Tipo

key

TEXT PRIMARY KEY

url

TEXT

prompt_hash

TEXT

output

TEXT

cached_at

INTEGER (ms epoch)

hit_count

INTEGER

last_hit_at

INTEGER

Hook SessionStart

En cada sesión nueva, el hook inyecta una estadística de una línea:

webcache: 142 pages cached, 38 hits, last fetch 3h ago

Omite la inyección si la caché está vacía.

TTL

Por defecto 7 días. Las entradas caducadas se eliminan en la siguiente lectura de la misma clave. Ejecuta una purga manual requiriendo src/cache.js y llamando a purgeExpired().

Límites

  • La clave de caché incluye el prompt -> diferentes prompts en la misma URL son entradas separadas. Elige prompts consistentes (p. ej., siempre "extraer título y contenido principal") para maximizar la tasa de aciertos.

  • La salida es lo que devuelva WebFetch (ya resumido por el modelo). La caché no lo vuelve a procesar.

  • Sin búsqueda semántica, sin embeddings. Solo coincidencia exacta de (url, prompt).

Licencia

MIT -- consulta LICENSE.

Available Tools

4 tools
cached_fetchA

Look up a URL+prompt pair in the local WebFetch cache. Returns cached output if present (instant), or "[CACHE_MISS] " if not. On CACHE_MISS, call WebFetch, then call cache_store with the result. Same URL+prompt across sessions hits the cache.

ParametersJSON Schema
NameRequiredDescriptionDefault
urlYesThe URL to fetch
promptYesThe prompt/instruction for the WebFetch

TDQS

A4.7/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Fully describes behavior: cache lookup, instant return on hit, cache miss string, and cross-session caching. Since no annotations are provided, the description carries the full burden and meets it.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Three sentences, front-loaded with purpose, each sentence adds value without redundancy. Fits the required structure.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Complete for a simple lookup tool: explains input, output, and fallback workflow. No output schema needed as return values are textual and explained.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Adds meaning beyond schema by explaining that url and prompt form a cache key and that prompt is an instruction for WebFetch. Schema coverage is 100%, but description enriches semantics.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Description clearly states it looks up a URL+prompt pair in a local cache, returning cached output or a cache miss message. This distinguishes it from sibling tools like cache_list, cache_stats, and cache_store.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Describes when to use (check cache) and provides a workflow on cache miss (call WebFetch then cache_store). However, it doesn't explicitly state when not to use or compare directly to siblings.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

cache_listB

List recently cached URLs (most recent first).

ParametersJSON Schema
NameRequiredDescriptionDefault
limitNoMax entries to return (default 50)

TDQS

B3.3/5.0
Behavior2/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations, the description carries the full burden of behavioral disclosure. It only states the output ordering but does not explain what constitutes 'recently cached', whether the list is global or scoped, or if there are any side effects (e.g., consuming cache entries). No safety or rate-limit information is provided.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single, front-loaded sentence that conveys the core purpose with no extraneous words. Every token is meaningful.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness3/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool's simplicity (1 optional param, no output schema, no nested objects), the description is minimally adequate. However, it could benefit from additional context such as whether the cache is global or per-user, or how 'recent' is defined, but the basic listing functionality is clear.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

The schema has 100% coverage for the single parameter 'limit', and the description adds no additional meaning beyond what the schema already provides. The baseline score of 3 is appropriate since the schema is sufficient.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description uses a clear verb 'List' and specifies the resource 'recently cached URLs' with ordering 'most recent first'. It distinguishes this from sibling tools like cached_fetch (which likely fetches a specific URL) and cache_store (which stores).

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines2/5

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 like cached_fetch or cache_stats. There is no mention of prerequisites, use cases, or exclusions. The user is left to infer 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.

cache_statsA

Return cache statistics: total entries, total hits, last cached timestamp.

ParametersJSON Schema
NameRequiredDescriptionDefault

No parameters

TDQS

A3.8/5.0
Behavior3/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

The description indicates a read-only operation (returning statistics) with no explicit destructive behavior. However, it lacks disclosure of potential side effects, authentication requirements, or rate limits. Since no annotations are provided, the description carries the full burden, and it only partially meets that need.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single, concise sentence that front-loads the purpose ('Return cache statistics') and lists the specific outputs. Every word is necessary and there is no redundancy.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the tool's simplicity (no parameters, no output schema), the description is sufficiently complete. It clearly lists the three statistics returned. It could optionally mention that the tool is safe to call at any time, but this is not required for basic completeness.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

The tool has zero parameters, so the description adds no parameter-specific information. According to guidelines, baseline score is 4 when there are 0 parameters, as no compensation is needed.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description explicitly states the tool returns cache statistics including total entries, total hits, and last cached timestamp. It uses a specific verb ('Return') and resource ('cache statistics'), and clearly distinguishes from sibling tools like 'cache_list' (which likely returns key listings) and 'cache_store' (which stores items).

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines2/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

No guidance is provided on when to use this tool versus its siblings ('cached_fetch', 'cache_list', 'cache_store'). The description only states what it returns, leaving the agent to infer appropriate usage without explicit context or exclusions.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

cache_storeA

Store a WebFetch result in the cache after a CACHE_MISS. Pass the original url, prompt, and the output text returned by WebFetch.

ParametersJSON Schema
NameRequiredDescriptionDefault
urlYes
outputYesThe output text returned by WebFetch
promptYes

TDQS

A3.9/5.0
Behavior2/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations, the description must disclose behavioral traits. It only states the storage action but does not mention what happens if the URL already exists (overwrite?), error cases, or authorization needs. This leaves significant ambiguity.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is concise, consisting of two sentences. The first sentence front-loads the purpose and trigger, and the second lists the parameters. No unnecessary words.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness3/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the lack of annotations and output schema, the description provides an adequate but minimal explanation. It identifies the tool's role in caching but omits details about idempotency, error handling, or response format, which would be helpful for a complete understanding.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is low (33%), but the description adds meaning by associating each parameter with its role: 'original url, prompt, and the output text returned by WebFetch.' This helps clarify the purpose of the url and prompt parameters beyond the schema.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool's action ('Store a WebFetch result in the cache'), the resource ('cache'), and the trigger condition ('after a CACHE_MISS'). It effectively distinguishes from sibling tools like cached_fetch, cache_list, and cache_stats by specifying its role in the caching workflow.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description explicitly specifies when to use the tool (after a CACHE_MISS), providing clear context. However, it does not mention when not to use it or offer alternatives (e.g., using cached_fetch for a subsequent lookup).

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

  1. 4 tool updatesv0.1.5
    • First observedcache_list
    • First observedcache_stats
    • First observedcache_store
    • First observedcached_fetch
Deploy Server
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

TDQS

A3.9/5.0

Scored across 4 tools

Disambiguation5/5

Each tool has a distinct function: lookup, list, stats, store. No overlap in purpose.

Naming Consistency4/5

All tools use 'cache_' prefix, but 'cached_fetch' uses past participle while others use 'cache_' as noun, a minor inconsistency.

Tool Count5/5

4 tools is well-scoped for a caching utility, covering essential operations.

Completeness4/5

Covers lookup, storage, listing, and stats, but lacks a clear operation to clear or invalidate cache entries.

Maintenance

ActivityMaintained
ResponsivenessUnresponsive

Resources

Related MCP Connectors

Related MCP Servers

  • ClaudeX

    kunwar-shah
    A
    license
    A
    quality
    C
    maintenance
    Persistent memory + FTS5 full-text search for Claude Code conversation history. Indexes ~/.claude/projects/ JSONL into SQLite, exposes 10 MCP tools (store/recall/search memories, browse sessions, get summaries) plus prompts. Includes a web UI for visual exploration
    10
    42 npm
    93
    MIT

  • sqlite-memory-mcp

    miles990
    A
    license
    B
    quality
    D
    maintenance
    Provides persistent memory, skill tracking, failure indexing, and context sharing for Claude Code using SQLite with FTS5 full-text search.
    23
    6 npm
    1
    MIT