Model Context Protocol server for ZeroBounce email validation, verification, and intelligence. Provides comprehensive email validation, AI scoring, email finder, domain search, and bulk processing capabilities for AI coding assistants.
The ZeroBounce MCP server has 15 tools with varying quality. Most tools have descriptions (13/15 visible), but many descriptions are brief and lack context for when to use them. Input schemas are present for all tools shown, with clear parameter types and required field markers. However, several critical issues reduce quality: (1) Tool names lack action verbs in some cases ('find_email' is good, but 'bulk_ai_scoring_send_file' is overly complex); (2) Parameter descriptions are often minimal (e.g., 'File ID to delete' lacks guidance on where to find a file ID or what happens after deletion); (3) Output schemas are not documented in the tool definitions, LLMs cannot predict what these tools return; (4) Error handling guidance is absent, tools like 'bulk_validate_delete_file' have destructive semantics but no recovery hints; (5) Several tools are inferred from test-mcp.js rather than explicit TypeScript definitions (bulk_validate_send_file, bulk_validate_file_status, bulk_validate_get_file, bulk_validate_delete_file, bulk_ai_scoring_send_file, bulk_ai_scoring_file_status, bulk_ai_scoring_get_file, bulk_ai_scoring_delete_file), capping their scores at 50. The server is functional but lacks production-grade polish.
Delete a bulk AI email scoring file from your account.
Get the status of a bulk AI email scoring file upload.
Get the results of a completed bulk AI email scoring file.
Send a CSV file for bulk AI email scoring. Returns a file ID for tracking.
Delete a bulk email validation file from your account.
Get the status of a bulk email validation file upload.
Output schemas not documented. Tool definitions provide input schemas but omit output/response schemas. LLMs cannot predict what fields to extract from responses, forcing guesswork on downstream tool calls.
Eight tools inferred from test-mcp.js rather than explicit TypeScript implementations. Tool definitions for bulk_validate_* and bulk_ai_scoring_* are not visible in src/tools/, only test file references exist. Cannot verify actual schema or error handling.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 55 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Get the results of a completed bulk email validation file.
Send a CSV file for bulk email validation. Returns a file ID for tracking.
Get the likely email address format for a domain or company (Domain Search API).
Find an email address by person name and domain or company name.
Get activity data for an email address to see if it has been active recently.
Get API usage statistics for a date range. Shows validation counts by status and sub-status.
Get the number of credits remaining in your ZeroBounce account.
Validate multiple email addresses in batch. More efficient for validating multiple emails at once.
Validate a single email address using ZeroBounce API. Returns detailed validation results including status, sub-status, and additional metadata.
Destructive operations lack recovery guidance and confirmation patterns. Tools like 'bulk_validate_delete_file' and 'bulk_ai_scoring_delete_file' have DESTRUCTIVE risk but no description warning or confirmation-request pattern to prevent irreversible accidents.
Parameter descriptions are minimal and lack actionable context. Examples: 'File ID to delete' does not explain where to obtain a file ID or what constitutes a valid format. 'Optional IP address of the user' lacks explanation of what IP accuracy improves (fraud detection, geo-fencing, etc.).
Tool names for bulk operations are verbose and lack clarity on sequencing. 'bulk_validate_send_file', 'bulk_validate_file_status', 'bulk_validate_get_file' should be shortened and made more discoverable (e.g., 'send_validation_batch', 'check_batch_status', 'fetch_batch_results').
No error handling documentation. No tool description mentions error conditions, recovery paths, or actionable error messages. LLMs will not know if a 'file not found' error is retryable or requires user action.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) present. Tools are marked with Risk levels in the spec but not exposed as structured annotations in the MCP protocol.
domain_search has no required parameters. Both 'domain' and 'company' are optional with no guidance on which to provide. LLMs will be uncertain whether both, one, or either is needed.