Model Context Protocol server for Apple Ads API integration, providing tools for campaign management, optimization, and reporting
Apple Ads MCP has 14 tools with schemas and descriptions present, but quality is inconsistent. Most tools have adequate naming (verb_noun pattern) and descriptions (50-150 chars), but parameter descriptions are sparse or missing. Several tools expose complex nested objects (payload parameters) with minimal guidance on required fields. Output schemas are not documented. Error handling is basic. The server reads from local files and makes API calls but lacks idempotency guarantees, confirmation steps for destructive operations, and clear recovery guidance.
Preview ad group creation without applying changes
Preview campaign creation without applying changes
Preview campaign deletion with confirmation text
Preview creative creation without applying changes
Preview keyword creation without applying changes
Preview keyword deletion with confirmation text
Get keyword suggestions for campaigns
Payload parameters (campaign_create_preview, creative_create_preview, shared_budget_create_preview) have type 'object' with additionalProperties=false but no documented required fields or nested schema. LLMs cannot infer what fields to include.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., policy IDs, campaign IDs) from responses.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2025-06-18+ | v2 |
Build optimization baseline from completed days of campaign data
Load bounded local optimization history entries
Build optimization plan with bounded actions; no receipt created
List all local optimization policies
Get a specific optimization policy with resolved balanced thresholds
Get server information and capabilities
Preview shared budget creation without applying changes
Destructive tools (keyword_delete_preview, campaign_delete_preview) require expectedText confirmation but lack guidance on what text to expect or how to obtain it. No dry-run or multi-step confirmation pattern.
keyword_suggestions accepts filters array with object items (field, operator, value) but provides no enum constraints, format guidance, or examples of valid operators. LLMs will hallucinate invalid filter syntax.
No error recovery guidance in tool descriptions. When a tool fails (e.g., policy not found, API error), LLMs have no hint about what to try next or whether the error is retryable.