Loading ToolBench
MCP server providing tools for Docker, Kubernetes, weather, user data, and math operations via stdio transport
This STDIO-only server has 13 tools with heterogeneous quality. Naming is inconsistent (verb-noun mostly present but some tools are domain-specific like 'math' rather than 'calculate_math'). Descriptions vary widely: some are adequate (20-50 chars), others generic or in Portuguese without clarity for English-speaking agents. Parameter descriptions exist but are sparse and often in Portuguese. Critical issue: no output schemas documented anywhere, tools return JSON-stringified responses with no structured definition of what fields agents should expect. No error handling guidance visible. Many parameters lack validation rules, ranges, or constraints. The 'math' tool enum uses Portuguese terms ('adição', 'subtração') which will confuse English-speaking LLMs. Kubernetes and Docker tools expose complex operations (set-context, start/stop containers) with minimal safety guidance. Overall, definitions are functional but sub-par for production agent reasoning.
Retornar os dados de usuário com score >= 950 e que estão ativos
Retornar a quantidade dos super usuários separados por país
Retornar o clima de uma cidade
Retorna as informações de um container docker
Retorna todos os contextos do kubernetes
Retorna o contexto atual do kubernetes
Retorna as pods do kubernetes
No output schemas documented for any tool. Tools return JSON.stringify() responses with no structured field definitions, making it impossible for LLMs to plan downstream tool calls or extract specific data. Pattern 'response-shaper' requires documenting return types.
Math tool enum uses Portuguese terms ('adição', 'subtração', 'multiplicação', 'divisão') instead of English. English-speaking LLMs will struggle to select valid enum values and may hallucinate English alternatives.
All descriptions are in Portuguese. For a globally-used MCP server, descriptions must be in English or both languages clearly separated. LLMs trained primarily on English may misinterpret Portuguese descriptions or fail to understand tool purpose.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 44 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 47 | - | v1 |
Busca logs de uma pod do kubernetes
Define o contexto atual do kubernetes
Retornar os containers criados na máquina
Realizar cálculos de soma/subtração/multiplicação/divisão de 2 números
Inicia um container docker
Para um container docker
No error handling guidance. Destructive tools (start-container, stop-container, kubernetes-set-context) lack confirmation steps, dry-run options, or error recovery paths. Pattern 'confirmation-request' requires agents to confirm irreversible operations.
Parameter descriptions are sparse and lack validation rules. 'containerId' descriptions offer no guidance on format or constraints. Kubernetes parameters like 'podName' and 'namespace' lack examples or constraints. Pattern 'constrained-input' requires explicit format/range documentation.
No pagination support on list tools. 'list-containers', 'kubernetes-get-contexts', 'get-super-users', 'get-top-countries' return raw results with no limit/offset/cursor parameters. Large result sets will exhaust context windows. Pattern 'paginated-result' requires limit and cursor/offset support.
Tools are missing permission declarations. No scope or permission metadata (e.g., 'read:docker', 'write:kubernetes') visible. Pattern 'scope-declaration' requires each tool to declare required permissions for audit and least-privilege agent configuration.