Payment execution layer for AI agents. MCP server + CLI. Charge, refund, escrow, split payments, metered billing, x402 enforcement, spend controls, dispute resolution, payment retry, and signed webhook event delivery for autonomous agents.
20 tools with complete schemas and descriptions, but significant gaps in parameter documentation and error handling guidance. Tool names follow verb_noun convention well (set_tool_price, create_customer, charge). Descriptions average ~80 chars, below the 194-char baseline for A+ tools. Most parameters lack detailed constraints (ranges, formats, dependencies). No documented output schemas. Error handling is minimal, tools do not guide recovery or categorize errors as retryable/user-fixable/fatal. Idempotency support exists (idempotency_key in charge) but not consistently applied. No confirmation patterns for irreversible operations (charge, fulfill_intent, release_escrow).
Charge a customer immediately. Creates and processes a payment in one step.
Register a new customer (agent or user) for payments.
Create an escrow that holds funds until a task between agents completes. Agent A funds escrow → Agent B does the task → Agent A releases. Solves agent-to-agent trust.
Create a payment intent — declares intent to pay before execution. Compatible with x402 protocol.
Generate x402 HTTP 402 payment requirements. Use when an agent requests a paid resource.
Fulfill a payment intent — execute the charge.
Look up a customer by ID.
No documented output schemas for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., what fields does charge() return? Does it include transaction_id, timestamp, confirmation_code?). Breaks tool chaining.
Irreversible operations (charge, fulfill_intent, release_escrow) lack confirmation/dry-run patterns. Agents can accidentally execute financial transactions without a safety gate. No recovery guidance in error responses.
Parameter descriptions lack constraints. 'amount' appears in 10+ tools but no description specifies range, precision (cents vs dollars?), or minimum/maximum. 'currency' enum is present but no description explains what happens if currency mismatches customer wallet. Ambiguous parameter semantics invite LLM errors.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 61 | 2026-07-28+ | v2 |
Check the status of an escrow.
Get details of a specific payment.
Generate a payment receipt with cryptographic signature.
Check the current pricing for a tool.
List payments with optional filters.
List all tool pricing.
Get payment analytics — total volume, success rate, revenue by tool.
Refund escrow funds back to the payer — call if the task was not completed or was rejected.
Refund a payment (full or partial). Credits back customer balance for internal payments.
Release escrow funds to the payee — call when the task is complete and you're satisfied.
Set pricing for an MCP tool or service. Use this to monetize your agent's tools.
Add prepaid balance to a customer account (internal ledger).
Verify that a payment succeeded. Returns validity status and transaction details.
No error handling guidance. Tools do not return actionable recovery messages. E.g., if charge() fails due to insufficient balance, does it return 'insufficient_balance' or a generic error? No categorization of retryable vs user-fixable errors.
Tool descriptions are terse (avg 70 chars vs 194-char baseline). Missing context on WHEN to use each tool. E.g., 'create_payment_intent' vs 'charge', when should an agent pick one over the other? No guidance on prerequisites or dependencies between tools.