battle-tested MCP server for AI agent payments and invoicing

every agent project that touches payments ends up re-implementing the same governance logic: spending caps, approval workflows, audit logs.

the missing piece is a standard MCP server that handles payments, invoicing, and reconciliation with policy enforcement built in.

i built mnemopay to be that server.

here's what it does:

• MCP-compatible tool interface — agents call standard MCP methods like propose_payment, list_invoices, reconcile_transaction
• policy enforcement — per-counterparty caps, approval thresholds, counterparty whitelists enforced before any transaction executes
• two-phase commit — the agent proposes a payment, the policy engine validates it, then the transaction commits or aborts atomically
• merkle audit trail — every proposed and executed transaction gets logged in a tamper-evident chain
• multi-rail support — wraps Coinbase x402, bank transfers, and other payment APIs with a unified governance layer

ran 672 tests across rollback scenarios, policy violations, and concurrent transaction attempts. the server handles race conditions and partial failures without leaving the payment state inconsistent.

the core insight: payment tooling for agents isn't just about calling an API — it's about enforcing policy constraints and maintaining an audit trail that satisfies EU AI Act Article 12 requirements.

if you're building agent infrastructure, you don't want to roll your own governance logic. use an MCP server that treats policy enforcement and auditability as first-class features.

mnemopay SDK v0.5.0 ships with the MCP server interface. you can plug it into any agent framework that supports MCP and get safe, auditable payments out of the box.