Loading ToolBench
An MCP server that acts as a hub for managing HTTP-based tools and backend MCP servers, with admin API for tool group and HTTP tool management
mcphub provides 17 tools with consistent naming (verb_noun pattern) and structured schemas. All tools have descriptions and input schemas with typed parameters. However, descriptions are minimal (10-50 chars, well below the 194-char baseline), parameter descriptions lack actionable constraints (format, range, enum values), and output schemas are not documented. Error handling is absent, no recovery guidance, no categorization of retryable vs fatal errors. Security concerns: no evidence of permission gates on destructive operations (delete_tool_group, delete_http_tool, delete_openapi_import). Composition is sound, tools are single-responsibility and chainable via IDs.
Create a new HTTP tool in a tool group
Create an OpenAPI import and generate HTTP tools from selected operations
Create a new tool group with HTTP tools configuration
Delete an HTTP tool
Delete an OpenAPI import and its associated tools
Delete a tool group
Get a specific HTTP tool
Get a specific OpenAPI import
Descriptions are critically short (10-50 chars vs 194-char baseline). Examples: 'List all configured tool groups' (32 chars), 'Get a specific tool group by ID' (31 chars). LLMs cannot distinguish when to use similar tools or understand prerequisites.
Parameter descriptions lack actionable constraints. 'request_timeout' has no format hint (e.g., '5s', '30000ms'). 'max_response_body_bytes' has no range (e.g., 1-10485760). 'method' has no enum constraint. LLMs will guess invalid values.
No output schemas documented. Tools return complex objects (toolGroupView, httpToolView with nested runtime, oauth, headers) but LLMs have no schema to parse responses. Breaks downstream tool chaining.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 58 | 2025-06-18+ | v2 |
Get a specific tool group by ID
Inspect an OpenAPI specification before importing tools
List audit events for tool groups and HTTP tools
List all HTTP tools in a tool group
List all OpenAPI imports for a tool group
List all configured tool groups
Probe a tool group to verify connectivity and configuration
Update an existing HTTP tool
Update an existing tool group
Destructive operations (delete_tool_group, delete_http_tool, delete_openapi_import) lack confirmation or dry-run support. No error handling guidance. Agents can permanently delete configurations without recovery path.
No error handling or recovery guidance in tool definitions. Errors return raw HTTP status codes (e.g., 'storage_failed', 'validation_failed') with no actionable next steps. LLMs cannot self-correct or retry intelligently.