XGuard — Agent Execution Gateway

Scoped agent execution. Server-side credentials, policy, budgets and verifiable receipts.

我该使用它吗

质量与安全性

B
描述质量
95%
模式完整度
64%
命名质量
80%
投毒风险
100%
权限匹配度
100%
协议合规性
100%

基于对工具定义和协议合规性的自动分析。

上下文开销

~988token 数(工具定义)
~775 B典型响应大小
对注意力有中等影响(占 128k 上下文窗口的 0.77%)

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": {
    "xguard-control-plane": {
      "url": "https://api.xguardgate.com/mcp"
    }
  }
}

远程端点

https://api.xguardgate.com/mcpstreamable-http

它能做什么

工具清单

工具(8)

🟢 只读🟡 写入🔴 删除⚪ 未知
🔴xguard_execute(intent, operation, capability, input, idempotency_key)

Execute an explicit scoped provider operation, or a supported public-source outcome. intent:demo is free. Never infer a mutation from ambiguous text.

输入模式

{
  "type": "object",
  "properties": {
    "intent": {
      "type": [
        "string",
        "object"
      ]
    },
    "operation": {
      "type": "string"
    },
    "capability": {
      "type": "string"
    },
    "input": {
      "type": "object"
    },
    "idempotency_key": {
      "type": "string"
    }
  },
  "additionalProperties": true
}
🔴xguard_secretless_call(capability, operation, input, target, method, ...)

Call an allowed provider operation with a capability. Enforces scope, call/credit budget and billing before server-side credential injection. Writes require idempotency_key.

输入模式

{
  "type": "object",
  "properties": {
    "capability": {
      "type": "string",
      "description": "Scoped XGuard capability, never an upstream key."
    },
    "operation": {
      "type": "string",
      "enum": [
        "github.repository.read",
        "github.issue.create",
        "github.issue.comment",
        "github.pull_request.create",
        "cloudflare.zone.read",
        "cloudflare.worker.metadata",
        "slack.channel.read",
        "slack.message.send",
        "notion.page.read",
        "notion.database.read",
        "notion.page.create",
        "notion.page.update",
        "openai.response.create",
        "anthropic.message.create",
        "gemini.content.generate",
        "stripe.customer.read"
      ]
    },
    "input": {
      "type": "object"
    },
    "target": {
      "type": "string",
      "format": "uri"
    },
    "method": {
      "type": "string",
      "enum": [
        "GET",
        "HEAD",
        "POST",
        "PUT",
        "PATCH",
        "DELETE"
      ]
    },
    "body_json": {},
    "idempotency_key": {
      "type": "string",
      "minLength": 8,
      "maxLength": 128
    }
  },
  "required": [
    "capability"
  ],
  "additionalProperties": false,
  "anyOf": [
    {
      "required": [
        "operation",
        "input"
      ]
    },
    {
      "required": [
        "target"
      ]
    }
  ]
}
🟢xguard_preflight(intent, operation, capability, input, idempotency_key)

Inspect validation, scope and remaining capability budget without reserving, billing or executing. Execution rechecks everything.

输入模式

{
  "type": "object",
  "properties": {
    "intent": {
      "type": [
        "string",
        "object"
      ]
    },
    "operation": {
      "type": "string"
    },
    "capability": {
      "type": "string"
    },
    "input": {
      "type": "object"
    },
    "idempotency_key": {
      "type": "string"
    }
  },
  "additionalProperties": true
}
🟢xguard_quote(intent, operation, capability, input, idempotency_key)

Inspect XGuard credit pricing for a capability, or obtain a signed x402 quote for a public outcome. Does not execute.

输入模式

{
  "type": "object",
  "properties": {
    "intent": {
      "type": [
        "string",
        "object"
      ]
    },
    "operation": {
      "type": "string"
    },
    "capability": {
      "type": "string"
    },
    "input": {
      "type": "object"
    },
    "idempotency_key": {
      "type": "string"
    }
  },
  "additionalProperties": true
}
🟢xguard_verify_receipt(proof, receipt, result_sha256)

Verify a ProofRail proof and optionally its bound x402 receipt and result digest. Signature validity does not prove source truth.

输入模式

{
  "type": "object",
  "properties": {
    "proof": {
      "type": "string",
      "maxLength": 16000
    },
    "receipt": {
      "type": "object"
    },
    "result_sha256": {
      "type": "string",
      "pattern": "^[a-f0-9]{64}$"
    }
  },
  "required": [
    "proof"
  ],
  "additionalProperties": false
}
🟢xguard_discover

Find the paid API catalog and buying instructions, plus explicit provider operations and input schemas.

输入模式

{
  "type": "object",
  "properties": {},
  "additionalProperties": false
}
🟢xguard_status

Read observed execution/MCP metrics and configuration; empty observations do not imply uptime.

输入模式

{
  "type": "object",
  "properties": {},
  "additionalProperties": false
}
🟢xguard_get_result(payment_identifier, quote)

Recover a paid public outcome with its payment identifier and original signed quote without execution or another payment.

输入模式

{
  "type": "object",
  "properties": {
    "payment_identifier": {
      "type": "string"
    },
    "quote": {
      "type": "string"
    }
  },
  "required": [
    "payment_identifier",
    "quote"
  ],
  "additionalProperties": false
}

社区

评价此服务器

证据

最近观测

已验证未记录版本8 个工具
已验证未记录版本3 个工具