Loading ToolBench

Reading the current corpus…

mcp-scorecard MCP server · C · ToolBench

mcp-scorecard

Grade any MCP server's agent-readiness in one command — stdio (10 protocol checks) AND hosted/remote servers (10 security + web agent-readiness checks). Open source, local-first, no credentials.

2026-07-28+STDIO1 tools

62/100

Definition Quality78

Protocol Readiness45

Supportability47

Spec posture

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

Static source inference · medium confidence · detected: stateful session

Deprecated protocol patterns detected

Summary

The 'audit' tool has a clear, action-oriented name and a comprehensive 194-character description that explains what it does, when to use it, and what it returns. The input schema is properly defined with a single required 'target' parameter (string type) and includes a meaningful description. However, the tool lacks explicit output schema documentation in the visible code, the description mentions '0 - 100 score, A - F grade, itemized checks, and actionable fixes' but no formal JSON Schema for the response structure is visible. Error handling appears robust (the tool validates npm packages, GitHub URLs, local paths, and https:// URLs), but recovery guidance is not documented in the tool definition itself. The tool follows single-responsibility principle (audit only) and uses verb_noun naming ('audit'). No parameter validation constraints (enums, ranges) are visible in the schema.

Tools (1)

auditread onlysource verified78/100

Grade an MCP server's agent-readiness and security. Accepts an npm package name, a GitHub repo URL, a local dist path (→ stdio protocol-quality checks), or a hosted https:// URL (→ web security + agent-readiness checks). Returns a 0–100 score, an A–F grade, itemized checks, and actionable fixes.

Issues (3)

HIGHtool ↗

Output schema not documented. Tool description mentions '0 - 100 score, A - F grade, itemized checks, and actionable fixes' but no formal JSON Schema for the response structure is visible in the code. LLMs cannot plan downstream operations without knowing the response shape.

audit

MEDIUMconstrained-input ↗

Input parameter 'target' lacks format constraints or examples. Description says 'npm package, GitHub URL, local dist path, or https:// hosted MCP server' but no enum, pattern, or validation rules are visible. LLMs may pass invalid formats.

audit

MEDIUMrecovery-guide ↗

No error recovery guidance in tool definition. The audit tool can fail (invalid package, unreachable URL, malformed repo) but the tool description does not explain what the LLM should do on failure or how to interpret error responses.

audit

Recommendations

Add a formal JSON Schema definition for the audit response (e.g. { type: 'object', properties: { score: { type: 'number', minimum: 0, maximum: 100 }, grade: { type: 'string', enum: ['A+', 'A', 'B', 'C', 'D', 'F'] }, checks: { type: 'array', items: { type: 'object' } }, fixes: { type: 'array', items: { type: 'string' } } } }) and include it in the tool definition.
Add format constraints to the 'target' parameter: describe valid npm package names (e.g. '@scope/name' or 'name'), GitHub URL patterns (e.g. 'https://github.com/owner/repo'), local paths (e.g. '/path/to/dist'), and https:// URL requirements.
Document error cases in the tool description: 'Returns an error if the target is unreachable, malformed, or not a valid MCP server. On error, the response includes a 'status' field set to 'error' and a 'message' field with recovery guidance.'
Consider adding a 'verbose' boolean parameter (default false) to control output detail level, reducing token waste for simple checks.
Add examples of valid targets to the parameter description (without including them in the description text itself, use a separate 'examples' field in the schema if supported).

Spec posture evidence

Inferred effective spec: 2026-07-28+.

Relies on Stateful initialize / Mcp-Session-Id (removed; protocol is stateless) - make each request self-contained

Score history

Overall score trend

First recorded score · v2 rubric

62/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	C	62	2026-07-28+	v2