Loading ToolBench
AI-powered email intelligence and analysis MCP server with IMAP/SMTP integration for email syncing, analysis, priority scoring, and response generation
Email Intelligence MCP server has significant definition quality gaps. Of 9 tools, only 4 have explicit schema definitions visible in source code (test_server, analyze_emails, get_priority_emails, send_email from tools/ directory). The remaining 5 tools (sync_mailbox, list_emails, get_email, search_emails, get_mailboxes) are defined inline in src/index.ts with minimal or no parameter descriptions. Tool descriptions range from adequate (test_server: 89 chars) to acceptable (list_emails: 64 chars, search_emails: 68 chars), but parameter-level documentation is almost entirely absent. No output schemas are documented for any tool. Error handling is non-existent, no recovery guidance, error categorization, or actionable messages visible. Security concerns are critical: src/tools files reference hardcoded credentials (search-skill-index.ts contains plain-text password 'calvin@kola!' and email 'calvin@kolaborate.co'), which violates pattern:secret-injection. No parameter descriptions for most inline-defined tools, forcing LLMs to guess intent.
Analyze emails to extract priority scores, categories, sentiment, action items, and generate response suggestions
Get full email details including body and analysis
List all known mailboxes and their sync state
Get high-priority emails requiring immediate attention
List email headers from the local database
Fast full-text search across all synced emails
Send an email message with optional attachments via SMTP
Trigger a background sync for a mailbox
Hardcoded credentials in source code (search-skill-index.ts contains plaintext email password)
5 of 9 tools have no parameter descriptions (sync_mailbox, list_emails, get_email, search_emails, get_mailboxes). LLMs cannot infer parameter meaning from names alone.
No output schemas documented for any tool. LLMs cannot determine what fields to expect or plan downstream tool calls.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 0 | <=2025-11-25 | v2 |
| 2026-03-09 | D | 53 | - | v1 |
Health check to verify the Email Intelligence MCP server is running
analyze_emails and get_priority_emails tools lack visible input schemas in source code, only names/descriptions are defined in tools/index.ts imports. Cannot verify schema completeness.
send_email and sync_mailbox tools have WRITE risk but descriptions do not explicitly state side effects. Tool descriptions must say 'modifies state' for write operations.
list_emails tool has complex nested filter parameter (unread, since, before, hasAttachment, category, minPriority) with no description of filter object structure or valid values.
No error handling guidance visible. Tools do not document recovery steps, error categorization, or actionable error messages. E.g., if search_emails returns no results, should LLM suggest sync_mailbox?
analyze_emails tool name does not follow verb_noun pattern clearly. Does it analyze existing emails or create analyses? Description says 'analyze emails' but no clarity on input (IDs? raw email text?).
get_email tool has boolean flags (includeBody, includeAnalysis) but no description explaining what these toggle or why an LLM would set them.
send_email tool is critical but no visible implementation or schema in src/index.ts code snippet. Only partial definition visible. Cannot assess completeness.