Loading ToolBench

Reading the current corpus…

nothumansearch MCP server · F · ToolBench

nothumansearch

A web crawler and indexing service that detects agentic-readiness signals (llms.txt, ai-plugin.json, OpenAPI specs, MCP servers, etc.) on websites and provides an API/web interface to search and monitor agent-ready services.

2026-07-28+STREAMABLE_HTTP1 tools

44/100

Definition Quality43

Protocol Readiness45

SupportabilityNot measured-

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence · evidence: Streamable HTTP

Current-spec patterns detected

Summary

NotHumanSearch is an MCP server that exposes a single tool: ProbeMCPJSONRPC. The tool has a basic description and input schema, but the definitions fall short of production quality. The description is adequate (143 chars) but lacks guidance on when to use the tool, what constitutes success, or how to interpret results. The input schema is minimal, a single string parameter with a description, but lacks format constraints, examples, or validation hints. No output schema is documented, leaving LLMs unable to plan downstream actions based on the response. Error handling is not specified, and there is no guidance on retryability or failure modes. The tool's naming is clear enough (verb_noun form, descriptive), but the overall design lacks the depth needed for confident LLM integration. The single-tool nature of this server means the average quality reflects that one tool's gaps.

Tools (1)

ProbeMCPJSONRPCread only50/100

POSTs a tools/list request to an MCP endpoint and verifies the response is valid JSON-RPC 2.0 with a result.tools array. This is the only way to confirm a claimed MCP server actually exists and responds.

Issues (5)

CRITICALtool ↗

No output schema documented. LLMs cannot understand what fields the tool returns, blocking downstream tool chaining and result interpretation.

ProbeMCPJSONRPC

HIGHconstrained-input ↗

Input parameter 'endpoint' lacks format constraints. No mention of required URL scheme (http/https), expected response time, or error cases when the endpoint is unreachable or returns invalid JSON-RPC.

ProbeMCPJSONRPC

HIGHtool-description ↗

Tool description does not explain when or why to call ProbeMCPJSONRPC instead of directly integrating with an MCP endpoint. Missing guidance on what constitutes a successful probe (e.g., does HTTP 200 suffice? Must tools array be non-empty?).

ProbeMCPJSONRPC

HIGHrecovery-guide ↗

No error handling or recovery guidance. Tool description does not specify what errors are possible (network timeout, invalid JSON-RPC, HTTP errors) or how to recover.

ProbeMCPJSONRPC

Recommendations

Document the complete output schema. Specify fields such as: success (boolean), tools (array of objects), error_message (string), response_time_ms (number). Include the structure of each tool object returned.
Add format constraints to the 'endpoint' parameter. Example: 'A fully qualified HTTP/HTTPS URL (e.g., https://example.com:8091). The endpoint must be accessible and running an MCP server.'
Expand the tool description to 200 - 300 characters. Include: (1) What the tool does (validates an MCP server responds correctly), (2) When to use it (to verify claimed MCP endpoints before integration), (3) What success looks like (HTTP 200 with valid JSON-RPC and non-empty tools array), (4) Typical failure modes (timeout, invalid JSON-RPC, connection refused).
Add explicit error handling guidance: 'If the endpoint does not respond within 10 seconds, returns non-200 HTTP, or sends invalid JSON-RPC, the tool will return success=false with an error_message field explaining the failure. Retry transient network errors; contact the endpoint owner for persistent failures.'
Specify parameter constraints: Add minimum/maximum response time, allowed HTTP methods (POST only), and authentication requirements if any.
Consider adding a dry_run or timeout parameter to allow the caller to control probe behavior (e.g., probe with a 5-second timeout vs. default 30 seconds).

Score history

Overall score trend

First recorded score · v2 rubric

44/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	44	2026-07-28+	v2