Tool-result size governor for MCP
Per-route output budgets for MCP: windowing, spill artifacts, signed omission receipts.
我该使用它吗
质量与安全性
A
基于对工具定义和协议合规性的自动分析。
上下文开销
~229token 数(工具定义)
~448 B典型响应大小
对注意力的影响极小(占 128k 上下文窗口的 0.18%)
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": {
"tool-result-size-governor": {
"url": "https://textforagents.com/products/tool-result-size-governor/mcp"
}
}
}远程端点
https://textforagents.com/products/tool-result-size-governor/mcpstreamable-http它能做什么
工具清单
工具(3)
🟢 只读🟡 写入🔴 删除⚪ 未知
⚪shape_result(route, args, text, result, fixture_tool, ...)
Shape one tool result under your per-route budgets -> shaped window + signed omission receipt. Trial: first 100 shaped results free once per operator; then $0.001/result (X-PAYMENT header) or a plan.
输入模式
{
"type": "object",
"properties": {
"route": {
"type": "string"
},
"args": {
"type": "object"
},
"text": {
"type": "string"
},
"result": {
"type": "object"
},
"fixture_tool": {
"type": "string"
},
"tool_args": {
"type": "object"
}
},
"required": [
"route"
]
}🟢fetch_slice(artifact_id, offset, limit)
Re-fetch a narrower slice of your own spilled payload, byte-exact, checksum re-verified. unknown_artifact is a clean error.
输入模式
{
"type": "object",
"properties": {
"artifact_id": {
"type": "string"
},
"offset": {
"type": "integer"
},
"limit": {
"type": "integer"
}
},
"required": [
"artifact_id"
]
}⚪verify_receipt(receipt)
Verify a signed omission receipt.
输入模式
{
"type": "object",
"properties": {
"receipt": {
"type": "object"
}
},
"required": [
"receipt"
]
}社区
证据
最近观测
已验证未记录版本3 个工具