Loading ToolBench
The server defines 5 tools with clear verb-based names and visible schemas using Zod validation. Tool descriptions are present and reasonably specific (average ~90 chars). However, parameter descriptions are sparse or missing, output schemas are not documented, and error handling lacks recovery guidance. The HTTP transport with OAuth adds credibility, but the definitions themselves fall short of production-grade standards: no input parameters have documented descriptions, no output schema is specified, and there is no guidance for error recovery or multi-step workflows.
Create a payment link with a specified amount. The amount must be a string with at least 4 characters, including a decimal point.
Get a list of all available payment methods from Mollie.
Get a list of the last transactions from Mollie, up to a certain limit.
Get information about the currently used Mollie profile.
Get all info for a given payment transaction. The transaction ID must start with "tr_"
Parameter descriptions are missing. The 'limit' param in fetch-transactions and 'amount' param in create-payment-link lack descriptions explaining valid ranges, formats, or constraints. LLMs cannot infer whether limit=100 is reasonable or amount='1.0' meets the '4 character minimum' rule.
Output schemas are not documented. Tools return JSON stringified responses, but there is no declared schema describing the returned fields, data types, or structure. LLMs cannot plan downstream tool chains or extract key fields (e.g., transaction IDs, payment links) without guessing.
No error handling guidance. Tool handlers catch errors but do not categorize them as retryable, user-fixable, or fatal. An LLM receiving 'Invalid token' has no guidance to re-authenticate; it receives a generic 500 response.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-21 | D | 59 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 46 | - | v1 |
Input schema for 'amount' parameter lacks actionable constraints. The description states 'minimum 4 characters' and 'including a decimal point' but does not clarify the currency format (e.g., EUR 1.00, or is USD 0.01 valid?). The Zod validator checks .includes('.').min(4), but this is not visible to the LLM in the description.
Empty input schemas for fetch-methods and profile-info lack any documentation that they are parameter-free. An LLM might attempt to pass parameters unnecessarily or be confused about when these discovery tools are appropriate.
No pagination or result-limit documentation. fetch-transactions accepts a 'limit' but there is no specification of the max value, default, or whether results are paginated. An LLM might request 10,000 transactions, exhausting the context window.