An advanced MCP server for comprehensive email operations including basic and advanced email management, contact management, and scheduling.
Email MCP Server has 19 tools with basic definitions, but exhibits significant quality gaps. Naming is mostly verb-prefixed and reasonable (send_email, read_recent_emails, search_emails), aligning with baseline conventions. However, descriptions are largely minimal and lack depth, most are 50-150 characters, below the 194-char production baseline. Parameter descriptions exist but are sparse: 'Recipient email address(es), separated by commas' is present but 'ID of the email to retrieve' lacks detail on format or what context it applies to. Schemas are present and typed (object with properties, required arrays), but lack constraints: no enums for boolean defaults, no regex patterns, no min/max bounds on numeric params. Output schemas are completely undocumented, no tool returns a shape or field list documented for LLM chaining. Error handling is absent from source: no recovery guidance, no actionable error messages visible. Security posture is weak: credentials (SMTP, IMAP config) must be injected server-side, but no evidence of vault/env var usage or secret-injection pattern in sample code. Composition is reasonable (each tool has one clear responsibility), but chaining support is weak due to missing output schemas and undocumented ID fields. Two tools (delete_email, delete_contact) are destructive with no dry-run or confirmation pattern. Overall, this server is in the D-to-low-C range: definitions exist but lack production polish and LLM-optimization.
Add a new contact to your contact list.
Send emails to multiple recipients.
Create an email draft for later editing or sending.
Delete a contact from your contact list.
Delete an email from the inbox.
Forward an existing email to new recipients.
Get all contacts in a specific group.
No output schemas documented for any tool. LLMs cannot infer return structure, breaking composition and chaining patterns. E.g., send_email likely returns a message_id, but this is not documented, so follow-up tools cannot safely reference it.
Destructive operations (delete_email, delete_contact) lack confirmation or dry-run patterns. No confirmation_request or idempotent_operation pattern. Agents could accidentally delete bulk records with no recovery path.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 45 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Get a specific email by its ID.
Get statistics about your email account.
List all contacts or search for specific contacts.
Mark an email as read or unread.
Read recent emails from the inbox.
Reply to an existing email.
Schedule an email to be sent at a later time.
Search contacts by name or email.
Search emails with advanced filters and pagination.
Send a simple email to one or more recipients.
Send an email with file attachments.
Update an existing contact's information.
Parameter descriptions are generic and lack format/constraint details. E.g., email_id parameter description says 'ID of the email to retrieve' but does not specify format (string UUID? IMAP UID? Gmail message_id?), making LLM-to-backend mapping error-prone.
Lists can grow unbounded, wasting context tokens.
Tool descriptions are too brief (40 - 55 chars typical, vs 194-char baseline). E.g., 'Get statistics about your email account' does not explain what metrics are returned, how to interpret them, or when to use this vs search_emails. LLMs lack selection context.
No error handling guidance in tool definitions. No recovery hints (e.g., 'If email not found, try search_emails() instead'), no error categorization (retryable vs user-fixable vs fatal), and no actionable error messages documented.
Credentials and secrets (SMTP server, username, password, API keys) must be injected via environment variables or vault. Source code does not show secret injection pattern or vault integration, this is a critical security gap for production deployment.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) visible in schemas. These are optional but recommended for guiding agent safety policies. Tools like delete_email and schedule_email should declare their intent.