Purpose: This is a domain search and naming engine based on the Model Context Protocol (MCP), intended to be called by MCP-compatible clients such as Claude, Codex, VS Code, Cursor, and Cline, for querying domain availability, prices, and naming suggestions.
Main capabilities: single-domain multi-TLD queries; batch queries of up to 100 names per single TLD; registrar price comparison; AI brand name suggestions (built-in fine-tuned Qwen 7B-DPO, no API Key required); social platform username availability checks; TLD metadata and restriction queries; detection of premium/auction domain signals; the name_project two-stage naming engine, which can score and rank candidate names and check domain, social, and npm availability.
Data sources: availability requires no key by default, prioritizing public RDAP, then the GoDaddy public endpoint (about 30 requests/minute, with circuit-breaker protection), and finally falling back to WHOIS. Price queries are an optional feature; it is recommended to configure the PRICING_API_BASE_URL backend (holding the Porkbun key), or use your own Porkbun/Namecheap key when no backend is configured.
Usage: can be run directly via npx, or built from source. Supports stdio (default, for Claude Desktop, Cursor, VS Code, etc.) and HTTP/SSE (for ChatGPT Actions, Web applications, LM Studio). HTTP mode provides endpoints such as /mcp, /api/tools/*, /openapi.json, /health, and /metrics. ChatGPT can create a Custom GPT and import the OpenAPI specification. By default, 12 tools are exposed; setting SLIM_TOOLS=true enables a slim set of 6 tools.
Limitations: prices require a configured backend or your own key, otherwise they are unavailable (availability still works normally); price and availability results are both from a single source at a single point in time, and must be verified on the registrar checkout page before purchase; premium/auction detection depends on the GoDaddy public endpoint and must be verified yourself; AI suggestions can point to a self-hosted Qwen inference endpoint, and fall back to the built-in offline semantic engine when not configured.
