KeepBal

Search, inspect and run agent tools with one key and one prepaid wallet.

我該用這個嗎

品質與安全性

A
說明品質
95%
結構描述完整度
80%
命名品質
100%
汙染風險
100%
權限相符程度
100%
協定合規性
100%

根據工具定義與協定合規性的自動化分析。

上下文成本

~579Token(工具定義)
~545 B典型回應大小
極小的注意力影響(128k 上下文的 0.45%)

這是每次將伺服器的工具載入模型上下文時所消耗的約略 token 數量。數量越高,可用於其他工作的注意力就越少。

安裝

一鍵安裝

將以下內容加入你的 `claude_desktop_config.json` 檔案:

{
  "mcpServers": {
    "keepbal": {
      "url": "https://keepbal.com/mcp"
    }
  }
}

遠端端點

https://keepbal.com/mcpstreamable-http

它能做什麼

工具清單

工具(4)

🟢 唯讀🟡 寫入🔴 刪除⚪ 未知
🟢search_tools(query, category, limit, offset)

Search live KeepBal tools and models locally, free of charge. Use 2–200 query characters. Results are ranked and paginated; use category models for models.

輸入結構描述

{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "minLength": 2,
      "maxLength": 200
    },
    "category": {
      "type": "string"
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 25,
      "default": 10
    },
    "offset": {
      "type": "integer",
      "minimum": 0,
      "default": 0
    }
  },
  "required": [
    "query"
  ],
  "additionalProperties": false
}
🟢inspect_tool(tool)

Read local public tool metadata, known parameters, pricing metadata and a JSON request example. Fixed tools show the per-call price; variable tools are billed at provider price + 35%. This is not an input-specific quote. Free; does not contact a provider. Unknown input schemas are explicitly marked.

輸入結構描述

{
  "type": "object",
  "properties": {
    "tool": {
      "type": "string",
      "minLength": 1
    }
  },
  "required": [
    "tool"
  ],
  "additionalProperties": false
}
🔴run_tool(tool, input)

Execute a public tool with your prepaid KeepBal wallet using the REST billing and access controls. Inspect first. Fixed tools charge the per-call price; variable tools are billed at provider price + 35%. Use input.max_cost_usd for a catalog spending cap; it must cover estimated cost and requires sufficient balance. Provider failures are not charged. Pending runs are polled for up to 25 seconds; use get_run to check later without starting another execution. Truncated or unsupported results must not be retried. Models use their documented REST endpoint.

輸入結構描述

{
  "type": "object",
  "properties": {
    "tool": {
      "type": "string",
      "minLength": 1
    },
    "input": {
      "type": "object",
      "additionalProperties": true
    }
  },
  "required": [
    "tool",
    "input"
  ],
  "additionalProperties": false
}
🟢get_run(run_id)

Check and settle an existing run by run_id through REST ownership and access controls. No new execution or additional polling fee; the original execution charge may settle. Use after run_tool reports pending.

輸入結構描述

{
  "type": "object",
  "properties": {
    "run_id": {
      "type": "string",
      "minLength": 1,
      "maxLength": 200
    }
  },
  "required": [
    "run_id"
  ],
  "additionalProperties": false
}

社群

為此伺服器評分

證據

近期觀測

已驗證未記錄版本4 個工具