Loading ToolBench
MCP server for Amazon SP-API with LWA-only authentication (no AWS credentials required)
Server has 10 tools with complete input schemas and descriptions. Naming follows verb_noun convention (get_, search_) appropriately. Descriptions are present and substantive (100-200 chars typical), meeting baseline expectations. However, output schemas are not documented, LLMs cannot predict response structure for downstream chaining. Parameters lack granular constraints (enums, min/max bounds). Error handling is basic (generic error wrapping). No tool annotations (readOnlyHint, idempotentHint) despite all tools being read-only. Missing dependency hints and recovery guidance in descriptions.
Get detailed product information from the Amazon catalog by ASIN. Returns product title, brand, category, BSR (Best Sellers Rank), images, bullet points, and attributes.
Get detailed FBA inventory information including breakdown of reserved quantities, unfulfillable reasons, and researching quantities. Provides deeper insight into inventory status.
Get financial event groups (payment disbursements) from Amazon Seller Central. Shows grouped financial events including fund transfers, account balances, and settlement periods. Useful for tracking payments and reconciliation.
Get financial events for a date range from Amazon Seller Central. Returns all financial events including sales, refunds, fees, reimbursements, and adjustments. Events are grouped by type for easy analysis.
Get FBA inventory summary including quantity available, reserved, inbound, and unfulfillable. Shows inventory health at a glance for all or specific SKUs.
Output schemas not documented. LLMs cannot predict response structure, field names, or types. Breaks tool chaining, agent cannot know what IDs to extract for downstream calls (e.g., does get_orders return order_id or orderId?).
No tool annotations (readOnlyHint, idempotentHint, destructiveHint). All 10 tools are read-only and idempotent, but LLMs cannot infer this. Annotations enable safe retry logic and reduce agent hesitation.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
Get detailed information about a specific Amazon order by order ID. Returns full order details including shipping address, buyer info, and order status.
Get all financial events for a specific Amazon order. Returns detailed financial breakdown including the sale amount, fees, taxes, shipping charges, and any adjustments for that order.
Get the line items (products) for a specific Amazon order. Returns details about each item including ASIN, SKU, quantity, price, and tax.
Retrieve a list of orders from Amazon Seller Central. You can filter by date range, status, and fulfillment channel. Returns order summaries including order ID, status, total, and buyer info.
Search the Amazon catalog by keywords or identifiers (ASIN, SKU, UPC, EAN). Returns a list of matching items with summaries. Useful for finding products or looking up items by identifier.
Pagination parameters (pageToken, nextToken) lack descriptions of token format and lifecycle. LLMs cannot reason about pagination without understanding token semantics.
Date parameters (postedAfter, createdAfter, etc.) lack format specification. Descriptions say 'ISO 8601' but do not clarify timezone handling or whether time component is required.
Error handling returns generic JSON error wrapper. No recovery guidance (e.g., 'Invalid ASIN format. ASINs are 10 alphanumeric characters.').