Blinka Machine Mouths
Paid text-to-schema extraction and failed-agent-workflow recovery over MCP.
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
~255token 数（工具定义）
~983 B典型响应大小
对注意力的影响极小（占 128k 上下文窗口的 0.20%）
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": {
"blinka-machine-mouths": {
"url": "https://blinka-machine-mouths.netlify.app/mcp"
}
}
}远程端点
https://blinka-machine-mouths.netlify.app/mcpstreamable-http它能做什么
工具清单
工具（2）
🟢 只读🟡 写入🔴 删除⚪ 未知
⚪structure_json(text, schema, instruction)
Paid $0.01 Base USDC. Convert unstructured text into a caller-supplied JSON Schema using only explicit source evidence. No URL fetch, private Mika state, or retained request body.
输入模式
{
"type": "object",
"properties": {
"text": {
"title": "Text",
"type": "string"
},
"schema": {
"additionalProperties": true,
"title": "Schema",
"type": "object"
},
"instruction": {
"default": "",
"title": "Instruction",
"type": "string"
}
},
"required": [
"text",
"schema"
],
"title": "structure_jsonArguments"
}⚪recover_workflow(goal, run_log, available_tools, constraints)
Paid $0.25 Base USDC. Recover a failed/stalled run into completed state, blockers, retry traps, unknowns, and the smallest safe next action constrained to exact caller-supplied tool names.
输入模式
{
"type": "object",
"properties": {
"goal": {
"title": "Goal",
"type": "string"
},
"run_log": {
"title": "Run Log",
"type": "string"
},
"available_tools": {
"anyOf": [
{
"items": {
"type": "string"
},
"type": "array"
},
{
"type": "null"
}
],
"default": null,
"title": "Available Tools"
},
"constraints": {
"default": "",
"title": "Constraints",
"type": "string"
}
},
"required": [
"goal",
"run_log"
],
"title": "recover_workflowArguments"
}社区
证据
最近观测
已验证未记录版本2 个工具