Loading ToolBench

Reading the current corpus…

clau MCP server · F · ToolBench

clau

Persona launcher for Claude Code — manages Claude Code personas with MCP server integration, environment variables, and keychain secrets

2026-07-28+STDIO14 tools

38/100

Definition Quality41

Protocol Readiness25

Supportability43

Spec posture

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

Static source inference · medium confidence

Current-spec patterns detected

Summary

This is a shell-script-based MCP server with 14 tools. Most tools lack proper input schemas, parameters are inferred from zsh argument parsing rather than explicitly defined JSON Schema. Descriptions exist but are often generic or incomplete. No output schemas are documented. The server manages MCP configurations and macOS keychain secrets, but the tool definitions are not formally registered with the MCP protocol; they are shell functions. This violates the core requirement that tools be registered with explicit schemas and descriptions.

Tools (14)

clauwritesource verified52/100

Launch Claude Code with a selected persona, optionally resuming a past session

clau --jsonread onlysource verified40/100

Output persona state as JSON including roots, scratch directory, and per-persona metadata

clau -hread onlysource verified40/100

Display the persona tree with origins (repo vs global), base layer indicators, and onboarding help

clau skillread only47/100

Print the clau persona authoring skill documentation

clau skill installwrite50/100

Install the clau persona authoring skill to the project or global .claude directory

clau-mcp addwrite50/100

Add an MCP server to one or more personas with stdio or http transport, environment variables, and headers

clau-mcp listread only50/100

List MCP servers configured for one or all personas

clau-mcp rmwrite50/100

Issues (5)

CRITICALtool ↗

Tools are not formally registered with MCP protocol. They are shell functions with inferred schemas, not explicit JSON Schema definitions. The MCP spec requires tools to be registered with name, description, and inputSchema.

clau-mcp addclau-mcp rmclau-mcp listclau-secret setclau-secret getclau-secret checkclau-secret auditclau-secret listclau-secret rmclauclau -hclau --jsonclau skillclau skill install

CRITICALresponse-shaper ↗

No output schemas documented. Tools return unstructured text or JSON without declaring the response structure. LLMs cannot plan downstream calls or extract fields reliably.

clau-mcp listclau-secret auditclau-secret listclau --json

HIGHtool-description ↗

Recommendations

Migrate from shell functions to a proper MCP server implementation (Node.js/TypeScript recommended). Register all 14 tools with explicit JSON Schema inputSchemas and documented output schemas.
Rename flag-based tools: 'clau -h' → 'describe_personas', 'clau --json' → 'export_personas_json'. Use verb_noun naming for all tools.
Add output schema documentation for all tools. E.g., clau-mcp list should document: {personas: [{name: string, servers: [{name: string, type: string, url?: string, command?: string}]}]}.
Expand parameter descriptions with format hints and constraints. E.g., 'env' should be 'Array of environment variable assignments as KEY=${VAR} pairs (e.g., ["API_KEY=${MY_API_KEY}"])'.
Implement structured error responses with recovery guidance. E.g., if persona not found, return 'Persona "dev" not found. Available personas: main, staging, prod. Did you mean "staging"?'

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