Model Context Protocol inspector - a tool for testing and debugging MCP servers with a web-based UI and CLI
MCP Inspector is a diagnostic tool for exploring MCP servers, not a production service tool. Its 8 tools have consistent naming (verb_noun pattern) and minimal but present descriptions. However, schemas are sparse, most tools have empty input objects or minimal parameter documentation. The tool set is narrow and introspective (list/read resources, prompts, tools; call tools; set logging), reflecting its purpose as an inspector rather than a domain-specific service. Parameter descriptions are absent for most inputs. Error handling and output schemas are not documented. This is a SUPPORT/DEBUG tool, not a user-facing service tool, which contextualizes the lower definition quality score.
Set the logging level for the MCP server
Get a prompt with optional arguments
List available prompts from an MCP server
List available resources from an MCP server
Read a specific resource by URI
List available resource templates from an MCP server
Call a tool with specified parameters
Output schemas are undocumented for all tools. Agents cannot reason about downstream data flow or what fields to extract from responses.
Error handling guidance is absent. Tools like tools/call and resources/read provide no recovery hints (e.g., 'Tool not found. Call tools/list to see available tools.') or error categorization (retryable vs fatal).
Parameter constraints are described in text but not formally in JSON Schema. The 'level' parameter in logging/setLevel lists valid values in description ('debug, info, notice, warning, error, alert, emergency') but schema lacks an enum array.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 64 | 2025-06-18+ | v2 |
| 2026-03-09 | F | 49 | - | v1 |
List available tools from an MCP server
tools/call and prompts/get accept 'args' as a free-form object with no schema validation. LLMs cannot determine what sub-fields are valid, required, or what type they accept.
logging/setLevel tool naming deviates from MCP conventions (camelCase setLevel vs snake_case set_level) and the tool itself reflects deprecated MCP patterns (server-initiated logging was removed in 2026-07).
Description lengths are borderline. Most descriptions (8/8 tools) are 34-70 characters, approaching the lower bound of the 10-1024 char baseline. While not critical for an inspector tool, richer context would aid selection (baseline median: 194 chars).