Server has clear, well-structured tool definitions with consistent naming (verb_noun pattern) and solid descriptions (avg 180 chars). All 5 tools have input schemas with type definitions and parameter descriptions. However, output schemas are undocumented, LLMs cannot predict response structure for planning. Error handling is minimal (generic HTTP errors). No tool annotations (readOnlyHint, idempotentHint). API key exposed as parameter violates secret-injection pattern. Missing pagination guidance despite list tools. Baseline: 549 production tools average 194 chars descriptions, 100% have documented returns, this server lacks return documentation entirely.

• Move _apiKey to server-side environment variable (e.g., LEADCONNECTOR_API_KEY). Remove from tool parameters. Use MCP server initialization to inject credentials, not per-call parameters.
• Document output schemas for all tools. Add to tool descriptions: 'Returns {contacts: [{id, name, email, phone, tags, customFields}], meta: {total, page}}' etc. Enable LLM to predict response structure.
• Add tool annotations in MCP server registration: set readOnlyHint=true for all 5 tools. Signals to agents these are safe to retry without side effects.
• Enhance error messages with recovery guidance. Catch 401 and return: 'Invalid API key. Verify in GoHighLevel Settings → Business Profile → API Key. If using Location key, ensure it matches the account scope.'
• Document pagination in list_contacts and list_opportunities descriptions: 'Supports limit (1-100, default 20). For large result sets, use limit + offset or implement cursor-based pagination via meta.nextCursor if available.'