blacktea

Spending controls for AI agents that pay online via x402. Set limits, require human approval, audit every payment.

Install

openclaw skills install @nmrtn/blacktea

blacktea: spending controls for paying agents

When to use

Use blacktea whenever this agent might pay for something online: a paywalled API, a premium data feed, any x402-enabled endpoint. blacktea sits between the agent and its wallet and enforces a spending policy before any money moves. It auto-approves small amounts, asks the human for larger ones, rejects what is over the line, and writes an audit log of every payment.

Setup (runs as a local MCP server)

blacktea ships as an MCP server. Register it once.

OpenClaw (~/.openclaw/openclaw.json, or openclaw mcp set):

json

{
  "mcp": {
    "servers": {
      "blacktea": {
        "command": "npx",
        "args": ["-y", "@nmrtn/blacktea-mcp"],
        "env": {
          "EVM_PRIVATE_KEY": "0x...",
          "BLACKTEA_POLICY": "/path/to/policy.json"
        }
      }
    }
  }
}

Hermes:

bash

hermes mcp add blacktea \
  --command npx \
  --args -y @nmrtn/blacktea-mcp \
  --env EVM_PRIVATE_KEY=0x... BLACKTEA_POLICY=/path/to/policy.json

No wallet handy? Add BLACKTEA_RAIL=mock (and optionally BLACKTEA_MOCK_AMOUNT=5) to exercise the full policy and approval flow with no x402 endpoint, no USDC, and no signing.

Tools this exposes

pay(url, intent, max_amount?): attempt a paid request. Runs the policy first. If the policy holds it for approval, returns status: "approval_required" with an intent_id and the amount, and does NOT pay yet. If allowed, settles and returns the response data plus a receipt.
approve_payment(intent_id): settle a held payment, after the human confirms.
reject_payment(intent_id): decline a held payment. Nothing is charged.
audit_query(limit?): recent payment events from the audit log.

Ask before spending

When pay returns approval_required, do NOT treat it as a failure. Tell the human the amount and what it is for, in plain language, and only call approve_payment after they say yes. Below the auto-approve limit it just pays. Over the hard limit it rejects. The human stays in the loop without leaving the chat.

Policy

A policy.json governs every payment. Example:

json

{
  "rules": [
    { "if": { "amount_lt": 1 }, "then": { "approve": true } },
    { "if": { "amount_gte": 100 }, "then": { "reject": "over_hard_limit" } }
  ],
  "default": { "approval": "callback" }
}

This auto-approves under 1 USDC, asks the human between 1 and 100, and rejects over 100. The full operator set is in the policy cookbook (see Links).

Pitfalls

The wallet key signs real payments. Use a dedicated agent wallet funded only with what you are willing to let the agent spend.
Through MCP, rely on the approve/reject tools, not a console prompt. The server holds the payment and asks for confirmation through the chat.

Verification

Run with BLACKTEA_RAIL=mock and BLACKTEA_MOCK_AMOUNT=5 and a policy that auto-approves only under 1 USDC. Ask the agent to pay any URL. It should pause, ask you to approve 5 USDC, settle only after you approve, and audit_query should then show the settled payment.