IPv4 network analysis and subnetting tutor with MCP integration. Provides tools for IP address analysis, subnet calculations, network diagram generation, and subnetting exercises.
IPMentor provides 4 networking tools with complete JSON schemas and reasonable descriptions. However, descriptions lack depth and guidance around when to use each tool, parameters have minimal context, and error handling is basic. Tool names are clear and action-oriented (ip_info, subnet_calculator, generate_diagram, generate_subnetting_exercise), but parameter descriptions are terse. Output schemas are partially documented but lack comprehensive field descriptions. The server follows basic tool definition patterns but falls short of production-grade LLM optimization: descriptions average ~80-90 chars (below the 194-char production baseline), parameter annotations are minimal, and there are no error recovery guides or dependency hints.
Generate network topology diagrams in PNG or SVG format showing subnets and their interconnections. Takes a network address and list of host counts per subnet.
Generate random subnetting exercises with complete solutions and network diagrams. Supports both equal subnet division and VLSM (Variable Length Subnet Mask) with variable host requirements.
Analyze IPv4 addresses with subnet masks. Returns IP in decimal and binary formats, subnet mask information, network address, broadcast address, first/last host, and total hosts.
Calculate subnets using different methods: max_subnets (divides network into N equal subnets), max_hosts_per_subnet (creates subnets with at least N hosts each), or VLSM (Variable Length Subnet Mask with custom host requirements per subnet).
Parameter descriptions are too terse and lack format/constraint guidance. E.g., 'subnet_mask' accepts '/24, 255.255.255.0, or 24' but the description does not explain these three formats or which is preferred. 'hosts_list' in generate_diagram has no guidance on ordering, range, or whether duplicates are allowed.
Tool descriptions lack context on WHEN and WHY to use each tool. 'ip_info' and 'subnet_calculator' both work with subnets but the description does not clarify which to call first or how they differ. Missing dependency hints (e.g., 'Call subnet_calculator first to get network addresses, then use generate_diagram to visualize').
Output schemas are not documented in tool descriptions. Callers cannot predict what fields are returned (e.g., 'analyze_ip' returns 10 fields including 'ip_binary', 'subnet_mask_binary', 'first_host', 'total_hosts', but none of these are advertised). This forces agents to discover via trial-and-error or inspect return values.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 54 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 32 | - | v1 |
'generate_subnetting_exercise' has no required parameters (only optional 'use_vlsm' boolean). The description does not explain what happens when called with no args, making it unclear whether a random exercise is always generated or if the tool requires setup.
Error handling returns generic error dicts (e.g., {'error': str(e)}) with no recovery guidance. If an LLM passes an invalid IP like '999.999.999.999', the error 'Invalid IP' tells the agent nothing actionable. Missing classification: is this retryable? Is it user-fixable? Should the agent ask for clarification?
'method' parameter in subnet_calculator is documented as an enum (max_subnets, max_hosts_per_subnet, vlsm) but the description does not explain the semantic difference. An LLM cannot infer from names alone: max_subnets divides a network into N equal subnets; max_hosts_per_subnet creates subnets with at least N hosts each. The 'number' parameter changes meaning depending on 'method'.
Tool descriptions lack information about what each tool returns and how results relate. E.g., generate_diagram requires 'ip_network' and 'hosts_list', but subnet_calculator returns a 'subnets' list with dicts containing 'subnet' and 'hosts' fields. The mapping between these structures is not documented, forcing the agent to infer or experiment.