Loading ToolBench
A Flask application that integrates with OpenAI's API to handle email-related tasks via MCP (Model Context Protocol)
This is not an MCP server, it is a Flask web application that acts as an MCP client. The app.py file contains no tool definitions, schemas, or MCP server implementation. It only instantiates an OpenAI client and passes an MCP server reference (via Mcp() object) to OpenAI's API. The actual Email tool is defined on a remote MCP server (MCP_SERVER_URL), which is not provided in this repository. Without access to the remote server's tool definitions, schemas, and descriptions, evaluation of tool quality is impossible. The single 'Email' tool listed is inferred from the TOOLS array configuration, not from explicit schema definitions in this codebase.
Email tool for handling email-related tasks from the user's own mailbox
This is a Flask client application, not an MCP server. No tool definitions, schemas, or input/output documentation exist in the codebase.
Tool schema is completely absent. The Email tool has no visible input schema, parameter types, or output structure definition.
Tool description is generic and lacks actionable detail. 'Email tool for handling email-related tasks from the user's own mailbox' does not explain what specific operations are supported (send, read, search, delete?) or when to use this tool.
No parameter documentation. Without seeing the remote MCP server implementation, it is impossible to verify that parameters have descriptions, type constraints, or validation guidance.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 25 | 2026-07-28+ | v2 |
Credentials (FASTMAIL_API_KEY, BEARER_TOKEN) are passed as HTTP headers in the MCP client configuration. If the remote server echoes these in responses or logs, they will enter the LLM context and be logged.