Remote MCP server for checking domain name availability using WHOIS and DNS resolution. Can be deployed to DigitalOcean App Platform.
The Domain Checker server has critical definition quality issues. Of the 5 listed tools, only 2 are truly distinct (check_domain and check_multiple_domains appear twice in the registry, and domain_info_resource is a resource handler not a tool). The source code shows only partial implementation. Descriptions exist but are generic (10-30 words). Input schemas are visible but lack detail on constraints, ranges, and accepted formats. The output format is entirely string-based ('✅ LIKELY AVAILABLE') rather than structured JSON, violating the response-shaper pattern. Error handling exists at the code level but does not guide LLM recovery. No enum constraints on domain input. No pagination or batching support. The server exposes internal method names (_check_whois, _check_dns_resolution) and implementation details in descriptions, rather than focusing on user intent.
Check if a single domain name is available for registration
Check if a single domain name is available for registration
Check availability for multiple domain names at once
Check availability for multiple domain names at once
Get domain availability information as a resource
Duplicate tool registration: check_domain and check_multiple_domains are listed twice in the tools array. This suggests either a copy-paste error in the manifest or a registry problem that will confuse LLMs.
No output schema documented. Tools return a formatted string (e.g., '✅ LIKELY AVAILABLE') rather than structured JSON. LLMs cannot extract fields like 'available', 'whois_available', 'dns_resolvable' for downstream reasoning or chaining. Violates response-shaper pattern.
Input parameter 'domain' lacks constraints. No validation that input is a valid domain name (e.g., regex pattern, length limits, TLD requirements). Description does not state format expectations. LLMs may pass malformed strings like 'not a domain' or 'http://example.com' without rejection.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 45 | <=2025-11-25 | v2 |
| 2026-03-09 | D | 52 | - | v1 |
Array input in check_multiple_domains lacks size constraints. No minimum (at least 1) or maximum (e.g., 100) limit specified. Description does not warn of batch limits. An LLM could attempt to check 10,000 domains in one call, causing timeout or resource exhaustion.
Descriptions are generic and do not explain WHEN to use each tool. 'Check if a single domain name is available' is bare, it does not explain the checking methods (WHOIS + DNS), what 'available' means (not registered + no A record), or dependencies on external services that may fail.
Error handling does not guide LLM recovery. The code catches exceptions and logs them, but tool descriptions do not document error cases or what to do if WHOIS or DNS lookups fail (e.g., 'If DNS times out, try again or check the domain manually'). Errors become dead-ends for the agent.
domain_info_resource is listed as a tool but described as a 'resource'. The MCP spec distinguishes resources (read-only files/content served by the server) from tools (executable functions). This tool should either be a resource with a proper URI scheme, or renamed to follow tool naming conventions (e.g., get_domain_info). Mixed categorization breaks tool discovery.
No idempotency or caching hint. Domain availability can change between calls, but the server does not document this. If an LLM calls check_domain twice on the same domain seconds apart and gets different results, it has no guidance on whether this is expected (e.g., domain just registered) or an error.
No timeout or rate-limit guidance. The code sets DNS resolver timeouts (5s) but does not document retry behavior, exponential backoff, or rate limits to external services (WHOIS, DNS). An LLM retrying in a loop could overwhelm upstream servers.