Loading ToolBench

Reading the current corpus…

mcpunit MCP server · F · ToolBench

← Browse

mcpunit

github.com/lee-to/mcpunit

Deterministic CI-first quality audit for MCP servers

2026-07-28+STDIO4 tools

38/100

Definition Quality43

Protocol Readiness32

Supportability35

Spec posture

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

Static source inference · medium confidence · detected: Logging

Deprecated protocol patterns detected

Summary

mcpunit is a quality audit tool for MCP servers, not a production MCP server itself. The bundled demo server (.mcp/demo.rs) contains 4 intentionally-flawed tools designed to trigger audit rules. Tool names are generic ('do_it'), descriptions are vague or missing context ('Helps with stuff'), and schemas are deliberately weak (additionalProperties: true). The demo is not meant for production use, it's a test fixture. Scoring reflects the demo's actual quality, which is poor by design.

Tools (4)

debug_payloadread onlysource verified25/100

Debug helper that accepts whatever input is available.

do_itirreversiblesource verified28/100

Helps with stuff.

exec_commanddestructivesource verified55/100

Execute an arbitrary shell command on the host machine.

write_filewritesource verified62/100

Write text content to any file path on disk.

Issues (5)

CRITICALtool ↗

Tool 'do_it' has generic, non-verb name that does not convey action. LLMs cannot infer intent from name alone.

do_it

CRITICALtool-description ↗

Tool 'do_it' description 'Helps with stuff' is vague and under 20 characters. Does not explain what the tool does, when to use it, or what it returns.

do_it

HIGHconstrained-input ↗

Tool 'debug_payload' accepts additionalProperties: true with no constraints. Allows arbitrary input, making it impossible for LLMs to reason about valid payloads.

debug_payload

HIGHpermission-gate ↗

Tool 'exec_command' executes arbitrary shell commands without documented safety constraints, scope hints, or permission gates. Marked DESTRUCTIVE but no error handling guidance provided.

exec_command

HIGHdefault-values ↗

Recommendations

Rename 'do_it' to a verb_noun pattern (e.g., 'process_target', 'execute_task') that clearly conveys the action.
Expand 'do_it' description to 50-200 characters explaining what it does, when to call it, and what it returns. Current 'Helps with stuff' provides zero signal.
Replace 'debug_payload' additionalProperties: true with explicit required fields and typed properties. Document which fields are expected and their purposes.
Add error handling guidance to 'exec_command' and 'write_file' descriptions: 'On failure, returns error code and stderr. Retryable: [yes/no]. Irreversible: [yes/no].'
Document the default value and safety implications of 'write_file.overwrite' parameter. Consider defaulting to false to prevent accidental overwrites.

Spec posture evidence

Inferred effective spec: 2026-07-28+.

Relies on Logging (deprecated) - log to stderr or use OpenTelemetry

Score history

Overall score trend

First recorded score · v2 rubric

38/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	38	2026-07-28+	v2