xrpl-mpp-payer¶
Install:
pip install xrpl-mpp-payer
xrpl-mpp --help
Install and run the optional payer agent server with:
pip install "xrpl-mpp-payer[mcp]"
xrpl-mpp mcp
The payer package wraps the client SDK for operators and agents. It provides a CLI, local auto-pay proxy, spend policy, and local receipt history.
Typical commands include:
export XRPL_MPP_EXPECTED_RECIPIENT=rYourApprovedMerchantAddress
xrpl-mpp pay https://merchant.example/premium --dry-run
xrpl-mpp proxy https://merchant.example --port 8787
xrpl-mpp receipts
xrpl-mpp budget
Use environment variables or a secret manager for the wallet seed, trusted network/RPC endpoint, allowed recipient and currency, and spend cap. Dry-run mode is the safest way to inspect a challenge before enabling signing.
Automatic signing fails closed until the operator supplies an approved
recipient with --recipient or XRPL_MPP_EXPECTED_RECIPIENT. The value is a
policy input, not a default copied from the challenge. Apply
XRPL_MPP_MAX_SPEND as a second independent ceiling for unattended use.
The close command sends a final cumulative voucher. It does not perform the
funder's on-ledger tfClose transaction or refund unused channel funding.
Agent integration¶
The payer's optional agent server is an operational wrapper around buyer
behavior. The standalone xrpl-mpp-mcp package is the protocol implementation
for native MPP metadata inside an MCP server. They solve different problems:
- use
xrpl-mpp-payerto give an agent a controlled payer capability; - use
xrpl-mpp-mcpto make an MCP operation itself payment-aware.
Never grant an agent an unlimited funded wallet. Apply recipient, currency, network, and spend limits outside model-generated arguments as well as inside the payer.
The payer agent tools do not expose recipient or spend-cap overrides to the
model. Non-dry-run pay_url, close_channel, and proxy_mode require
XRPL_MPP_EXPECTED_RECIPIENT and XRPL_MPP_MAX_SPEND in operator-controlled
process configuration. The environment value is an independent ceiling and
cannot be raised by a tool call.