Reclypse Utility Node
Five text and JSON utilities. Paid execution: 0.01 USDC standard or 0.05 USDC priority on Base.
我该使用它吗
质量与安全性
基于对工具定义和协议合规性的自动分析。
上下文开销
这是每次将服务器的工具加载到模型上下文窗口时所消耗的大致 token 数。数值越高，可用于其他任务的注意力就越少。
安装
一键安装
将以下内容添加到你的 `claude_desktop_config.json` 文件中：
{
"mcpServers": {
"utility-node": {
"url": "https://a2a.reclypse.ai/mcp"
}
}
}远程端点
https://a2a.reclypse.ai/mcpstreamable-http它能做什么
工具清单
工具（5）
⚪sanitize_text(raw_input)
Deterministic Unicode/control-noise normalization and selected instruction-pattern flags. Does not guarantee prompt-injection removal or make untrusted text safe. Retail prices and enabled payment streams are declared in server metadata; a valid payment is required for execution.
输入模式
{
"type": "object",
"properties": {
"raw_input": {
"type": "string",
"maxLength": 65536
}
},
"required": [
"raw_input"
],
"additionalProperties": false
}🟢structure_json(text, schema_target)
Parse JSON or a complete JSON code fence. Default preserves non-JSON text without inventing fields; object requires an object and records requires an array of objects. Rejects duplicate keys/non-finite values. Retail prices and enabled payment streams are declared in server metadata; a valid payment is required for execution.
输入模式
{
"type": "object",
"properties": {
"text": {
"type": "string",
"maxLength": 65536
},
"schema_target": {
"type": "string",
"enum": [
"default",
"object",
"records"
],
"default": "default"
}
},
"required": [
"text"
],
"additionalProperties": false
}🟢clean_markdown(content)
Extract text from HTML while ignoring script/style content and normalize documented formatting artifacts. Does not promise arbitrary malformed-Markdown repair or HTML security validation. Retail prices and enabled payment streams are declared in server metadata; a valid payment is required for execution.
输入模式
{
"type": "object",
"properties": {
"content": {
"type": "string",
"maxLength": 65536
}
},
"required": [
"content"
],
"additionalProperties": false
}🟢chunk_document(document, max_chunk_size)
Prefer paragraph/sentence boundaries, then split deterministically. Concatenated chunks reproduce the input exactly. Limits measure Unicode code points, not model tokens. Retail prices and enabled payment streams are declared in server metadata. Legacy prepaid credits use separate payload quotes.
输入模式
{
"type": "object",
"properties": {
"document": {
"type": "string",
"minLength": 1,
"maxLength": 65536,
"description": "Non-whitespace document, at most 64 KiB of UTF-8."
},
"max_chunk_size": {
"type": "integer",
"minimum": 1,
"maximum": 16000,
"default": 1000
}
},
"required": [
"document"
],
"additionalProperties": false
}⚪clean_and_chunk(document, max_chunk_size)
Compound utility: Cleans markdown artifacts and semantically chunks documents in one atomic transaction. Retail prices and enabled payment streams are declared in server metadata. Legacy prepaid credits use separate payload quotes.
输入模式
{
"type": "object",
"properties": {
"document": {
"type": "string",
"minLength": 1,
"maxLength": 65536,
"description": "Non-whitespace document, at most 64 KiB of UTF-8."
},
"max_chunk_size": {
"type": "integer",
"minimum": 1,
"maximum": 16000,
"default": 1000
}
},
"required": [
"document"
],
"additionalProperties": false
}社区
证据