MCP Registry for Bankr Industrial Capability Network. Search 360+ AI agent capabilities: DeFi risk, customs, logistics, compliance, manufacturing, finance. Each capability is payable via x402 (USDC on Base).
Server has 5 tools with basic schemas and descriptions, but significant gaps limit production readiness. All tools are READ_ONLY and follow verb_noun naming (search_, get_, list_). Descriptions exist but are generic (avg ~120 chars, below 194-char baseline). Parameters lack type specificity, many use string without enums or constraints. Output schemas are undocumented. Error handling is absent. Tool #4 (list_tasks) and #3 (list_categories) have unused/dummy parameters (_unused), signaling incomplete design. No pagination guidance despite search_capabilities returning 361 items. Inferred task/category/gain logic via helper functions suggests tool definitions may be partially generated rather than explicitly designed.
Get full details of a specific Bankr x402 capability by name.
Get capability cards formatted for X.com social media posts. Returns batch of cards ready for publishing.
List all capability categories with counts.
List all task types with capability counts.
Search Bankr x402 capabilities by task, category, or keyword. Returns structured cards with: cap, task, when, gain, pay, cost, hub. 361 capabilities available for AI agents.
Output schemas undocumented. No tool returns a documented schema, LLMs cannot predict response structure, forcing trial-and-error parsing. search_capabilities returns 'structured cards' but fields are inferred from code (cap, task, when, gain, pay, cost, hub, description, category). get_capability returns 'full details' with no schema. list_categories and list_tasks return counts but structure is opaque.
Dummy/unused parameters reduce clarity. list_categories and list_tasks both accept '_unused' string parameter marked optional with description 'Not used'. This signals incomplete API design and confuses LLMs about whether the parameter is required or has side effects.
No pagination or result-limiting guidance. search_capabilities accepts 'limit' (default 20) but no documentation of max results, total count, or next_cursor. With 361 capabilities available, LLMs may request unbounded results or retry without knowing when to stop.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 47 | 2026-07-28+ | v2 |
Parameter constraints missing. 'category' parameter lists 11 valid values in description text ('defi_crypto, customs_logistics, ...') but not as an enum. LLMs may hallucinate invalid categories. 'task' parameter similarly lacks enum constraint despite inferTask() function defining a fixed set.
No error handling or recovery guidance. Tools have no documented error cases, retryability, or actionable error messages. If search_capabilities finds no results or get_capability receives an invalid name, the LLM has no guidance on what to do next.