Loading ToolBench
MCP server for Governance, Risk, and Compliance operations
Server has 6 tools with basic schemas and descriptions, but significant gaps prevent production use. Tool names follow verb_noun convention (good), but descriptions are minimal (avg 30 chars, well below 194-char baseline). Input schemas present but lack depth: no enums for constrained fields (e.g., 'status' in createComplianceReport), no output schemas documented, no error handling guidance. Parameters have types but descriptions are sparse. No pagination support for list tools. Security: no visible auth validation implementation despite bearer token declaration. Composition: list tools return empty schemas, making chaining impossible.
Create a new compliance report
Create a new policy document
Create a new risk assessment
Get all compliance reports
Get all policy documents
Get all risk assessments
List tools have empty input schemas ({}) and no documented output schemas. LLMs cannot understand what fields to expect or how to chain results to other tools.
Constrained fields lack enums: 'status' in createComplianceReport accepts free-form strings instead of enum ['compliant', 'non-compliant', 'in-progress']. LLMs will hallucinate invalid values.
Tool descriptions are 20-35 characters (well below 194-char baseline). 'Create a new risk assessment' does not explain WHEN to use it, what it returns, or prerequisites. LLMs cannot reliably select the right tool.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 38 | 2026-07-28+ | v2 |
No output schemas documented. Handlers return unknown types. LLMs cannot plan downstream calls or extract required fields (e.g., riskId for follow-up updates).
No error handling or recovery guidance. Handlers lack try-catch, validation, or actionable error messages. LLMs receive raw errors with no guidance on retry or alternatives.