treasury-risk-mcp

MCP server for the Treasury Risk Analyzer: deterministic financial decision services for AI Treasury and CFO agents. It covers treasury risk, FX exposure, the liquidity forecast, and one question: "Can we safely execute this payment today?"

Each call is paid in USDC on Base via x402. There is no account, no API key and no subscription. Your agent pays per call from its own wallet, and only when the input is valid.

Tools

No LLM computes any figure: every metric is deterministic and documented in the response's methodology.

Related MCP server: Hermes Plant — Agent Commerce Assurance

Install

Claude Desktop (claude_desktop_config.json), Cursor, or any MCP client:

{
  "mcpServers": {
    "treasury-risk": {
      "command": "npx",
      "args": ["-y", "treasury-risk-mcp"],
      "env": {
        "X402_PRIVATE_KEY": "0x…",
        "MAX_PAYMENT_USD": "2.00"
      }
    }
  }
}

Claude Code:

claude mcp add treasury-risk -e X402_PRIVATE_KEY=0x… -e MAX_PAYMENT_USD=2.00 -- npx -y treasury-risk-mcp

Wallet safety: use a dedicated wallet that holds only a few USDC. The key stays in the MCP process on your machine. It is only used to sign an EIP-3009 USDC transfer for the exact price the service asks for, never more than MAX_PAYMENT_USD. No ETH is needed for gas, because the facilitator submits the transfer.

How payment works

1. The tool sends your input to the API. Invalid input gets a free 400 and you are never charged.

2. Valid input gets 402 Payment Required with the price.

3. The server signs a USDC authorization for that price and retries.

4. The payment settles on Base, and the analysis comes back with a _payment receipt (transaction hash, payer, network).

Limits

• The output is analytics and decision support, not investment advice. It does not execute any payment.

• It only analyses the data you supply. It does not fetch market data.

• The HTTP API is also callable directly. Its OpenAPI spec is at /openapi.yaml and /openapi.json.