← 返回
开发者工具 Key 中文

Kash - Agentic Payment Provider

Pay for APIs, tools, and services from your agent's Kash wallet. Spends below $5 are autonomous; above $5 requires explicit user YES. Requires KASH_KEY and K...
使用智能体的Kash钱包支付API、工具和服务。低于5美元自主支付,超过5美元需用户确认。需要KASH_KEY和K...
devfaraaz devfaraaz 来源
开发者工具 clawhub v1.0.7 1 版本 100000 Key: 需要
★ 2
Stars
📥 675
下载
💾 4
安装
1
版本
#latest

概述

Kash Payment Skill

This skill gives your OpenClaw agent access to a Kash wallet so it can pay for external services autonomously, within your configured budget.

Security model

  • KASH_KEY and KASH_AGENT_ID are required. The skill will fail at load time if either is missing — it will not silently proceed.
  • KASH_API_URL is validated against an allowlist (api.kash.dev and localhost only) at startup. Setting it to any other domain is rejected immediately to prevent KASH_KEY from being sent to an untrusted server.
  • KASH_BUDGET is enforced locally in code as a session cap. It is not just a guideline — the spend function checks it before every call.
  • Spends above KASH_SPEND_CONFIRMATION_THRESHOLD ($5.00 default) require confirmed=true, which the agent must only set after receiving an explicit YES from the user in the current conversation.
  • Budget enforcement happens at two layers: locally (KASH_BUDGET) and server-side (Kash dashboard budget). Both must pass. The server is the authoritative source of truth.

Tools provided

kash_spend

Spend from the Kash agent wallet before making a paid API call.

Parameters:

  • amount (number, required) — amount in USD
  • description (string, required) — what you are paying for
  • merchant (string, optional) — name of the service
  • confirmed (boolean, optional) — set true only after explicit user YES for spends above threshold

Return values:

  • OK. Spent $X for "..." — spend succeeded
  • CONFIRMATION_REQUIRED: ... — ask user for YES, then retry with confirmed=true
  • LOCAL_BUDGET_EXCEEDED: ... — session cap hit, stop and notify user
  • BUDGET_EXCEEDED: ... — server-side budget hit, stop and notify user
  • AGENT_PAUSED: ... — agent paused by user in Kash dashboard
  • UNAUTHORIZED: ... — KASH_KEY invalid or expired
  • ERROR: ... — unexpected failure

kash_balance

Check remaining budget without spending. Returns both server-side balance and local session cap.

When to use this skill

Use kash_spend BEFORE making any paid external call — API calls, web searches, data purchases, or any service that charges per request. Always call it before the paid operation, not after.

Use kash_balance before starting a multi-step task that will require several paid operations.

Rules the agent must follow

  1. Always call kash_spend BEFORE the paid call, never after
  2. If CONFIRMATION_REQUIRED is returned, ask the user for explicit YES — never bypass it
  3. If BUDGET_EXCEEDED or LOCAL_BUDGET_EXCEEDED is returned, stop the task immediately and tell the user
  4. Never set confirmed=true without a real user confirmation in the current conversation
  5. Never attempt to work around a budget limit
  6. If kash_spend returns any error, do NOT proceed with the paid call

Kash Payment Skill

Use this skill to pay for external APIs, tools, and services from your Kash agent wallet.

Setup

Configure in ~/.openclaw/openclaw.json:

{
  "skills": {
    "entries": {
      "kash": {
        "enabled": true,
        "apiKey": "ksh_live_...",
        "env": {
          "KASH_KEY": "ksh_live_...",
          "KASH_AGENT_ID": "your-agent-id",
          "KASH_BUDGET": "50",
          "KASH_SPEND_CONFIRMATION_THRESHOLD": "5.00"
        }
      }
    }
  }
}

Get KASH_KEY and KASH_AGENT_ID from kash.dev/dashboard/agents after registering your agent.

Tools

kash_spend

Call this BEFORE any paid API call or service. Never after.

Parameters:

  • amount (number, required) — cost in USD e.g. 0.003
  • description (string, required) — what you are paying for e.g. "serper web search"
  • merchant (string, optional) — service name e.g. "Serper"
  • confirmed (boolean, optional) — set true ONLY after explicit user YES for spends above threshold

Returns:

  • OK. Spent $X for "..." — proceed with the paid call
  • CONFIRMATION_REQUIRED: ... — ask user for YES, then retry with confirmed=true
  • LOCAL_BUDGET_EXCEEDED: ... — stop, tell user to top up at kash.dev
  • BUDGET_EXCEEDED: ... — stop, tell user to top up at kash.dev
  • AGENT_PAUSED: ... — tell user to resume agent at kash.dev/dashboard/agents
  • UNAUTHORIZED: ... — tell user their KASH_KEY may be invalid
  • ERROR: ... — do not proceed with the paid call

kash_balance

Check remaining budget. Use before starting multi-step paid tasks.

No parameters required.

Rules

  1. Always call kash_spend BEFORE the paid call, never after
  2. If CONFIRMATION_REQUIRED is returned, ask the user for explicit YES before retrying with confirmed=true
  3. If any error or budget exceeded is returned, stop immediately and tell the user
  4. Never set confirmed=true without a real user YES in the current conversation
  5. Never attempt to work around a budget limit
  6. Never reveal KASH_KEY value in any message or log

版本历史

共 1 个版本

  • v1.0.7 当前
    2026-03-29 21:38 安全 安全

安全检测

腾讯云安全 (Keen)

安全,无风险
查看报告

腾讯云安全 (Sanbu)

安全,无风险
查看报告

🔗 相关推荐

ai-agent

Self-Improving + Proactive Agent

ivangdavila
自我反思+自我批评+自我学习+自组织记忆。智能体评估自身工作、发现错误并持续改进。
★ 1,383 📥 320,864
ai-agent

Skill Vetter

spclaudehome
AI智能体技能安全预审工具。安装ClawdHub、GitHub等来源技能前,检查风险信号、权限范围及可疑模式。
★ 1,230 📥 268,172
ai-agent

self-improving agent

pskoett
捕获经验教训、错误及修正内容,以实现持续改进。适用于以下场景:(1)命令或操作意外失败;(2)用户纠正Claude(如“不,那不对……”“实际上……”);(3)用户请求的功能不存在;(4)外部API或工具出现故障;(5)Claude发现自身
★ 4,086 📥 813,737