MCP server for NetBox infrastructure management system providing tools to query and manage network infrastructure data
NetBox MCP has 4 tools with basic schemas and descriptions, but significant gaps limit production readiness. All tools have descriptions (10-100 chars, mostly adequate) and input schemas with type info. However, parameter descriptions are minimal or generic ('dict of filters'), output schemas are undocumented, and error handling is absent. No tool annotations (readOnlyHint, destructiveHint). Tools lack pagination despite querying potentially large datasets. Parameter validation is weak, object_type accepts any string and validates at runtime with a generic error. No guidance on filter format or valid filter keys. Descriptions do not explain when to use each tool (get_objects vs get_count_objects distinction is unclear). Overall structure is functional but falls short of production-grade agent tool patterns.
Count objects in NetBox based on their type and filters
Retrieve custom fields for NetBox objects
Get detailed information about a specific NetBox object by its ID
Get objects from NetBox based on their type and filters
Output schemas undocumented. Tools return raw API responses without documented field structure, forcing LLMs to infer result shape and breaking downstream tool chaining.
Parameter 'filters' lacks schema detail. Described as 'dict of filters' with no enum, format, or example. LLMs cannot determine valid filter keys or structure without NetBox API knowledge.
No pagination support. get_objects and get_custom_fields query potentially large datasets but lack limit, offset, or page parameters. Results could exceed context window.
Error handling is silent. get_count_objects catches exceptions and returns 0; get_objects has no visible error handling. LLMs receive no guidance on failure causes or recovery steps.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | <=2025-11-25 | v2 |
No tool annotations. All tools are read-only but lack readOnlyHint annotation. Agents cannot infer safety properties without explicit metadata.