MCP documentation

The main tool. Checks one or more labels against a set of TLDs and sorts every combination into available or taken. Availability comes from a registry-derived list of every registered domain, refreshed daily, so a single call covers every extension at once rather than issuing a WHOIS lookup per name. It answers in milliseconds even across the whole catalogue. Leave `tlds` out and it uses the caller's favorites, falling back to the most popular extensions for a new account.

{
  "name": "search_domains",
  "arguments": {
    "names": ["northbeam"],
    "tlds": ["com", "io", "dev"]
  }
}

{
  "results": [
    {
      "name": "northbeam",
      "available": ["dev"],
      "taken": ["com"]
    }
  ],
  "tldsChecked": 3,
  "tldSource": "requested"
}

The other half of picking a name: a domain is only really available if the handles are too. Checks X, YouTube, TikTok, Snapchat, Pinterest, Discord, GitHub, Telegram and Twitch. Unlike domain search there is no dataset behind this — no platform publishes its username list — so each platform is probed live through a rotating proxy pool and the answers are cached. Results are split four ways, and the last two matter: `unknown` means the check could not be completed (the platform blocked it or timed out) and must never be reported as free, while `invalid` means that platform's own username rules reject the string outright. Checking several names at once skips X, which bans on volume.

{
  "name": "check_usernames",
  "arguments": {
    "usernames": ["northbeam"]
  }
}

{
  "results": [
    {
      "username": "northbeam",
      "available": ["twitch", "telegram"],
      "taken": ["x", "github", "youtube"],
      "unknown": [],
      "invalid": []
    }
  ]
}

Applies the deterministic naming generators to a keyword and returns candidates by category: brandable, multiword, short, creative, modern, classic, professional and synonyms. No model is involved, so the same keyword always returns the same list. Each candidate arrives with availability already resolved against a short extension list; pass anything promising to search_domains for the full catalogue.

{
  "name": "generate_domains",
  "arguments": {
    "keyword": "orbit",
    "categories": ["brandable"],
    "count": 3
  }
}

{
  "keyword": "orbit",
  "tldsChecked": ["com", "io", "dev"],
  "candidates": [
    { "domain": "orbitly",  "category": "brandable",
      "available": ["io", "dev"] },
    { "domain": "orbitify", "category": "brandable",
      "available": ["com", "io", "dev"] },
    { "domain": "orbitfy",  "category": "brandable",
      "available": ["com", "dev"] }
  ]
}

Returns the favorite TLDs saved on the calling account, in the order they were saved. Call it before search_domains to scope a check to the extensions this person actually cares about, or leave search_domains to do it, which is what it does when you pass no `tlds`. An empty list can also mean the person keeps their extensions in their browser rather than on the account — that copy is out of reach here.

{
  "name": "list_favorite_tlds",
  "arguments": {}
}

{
  "tlds": ["com", "io", "dev", "co"],
  "total": 4,
  "limit": 200
}

Saves one TLD to the calling account. Every later search checks favorites first and shows them above the rest of the grid. Adding one that is already saved comes back as an error rather than silently doing nothing, so an assistant can tell the difference. If the person has "Save my favorite TLDs" switched off, their extensions live in their browser only: nothing is stored and the answer says so.

{
  "name": "add_favorite_tld",
  "arguments": { "tld": ".ai" }
}

{
  "added": "ai",
  "total": 5,
  "limit": 200
}

Removes one TLD from the calling account. Removing one that is not saved comes back as an error, so a mistaken call is visible rather than silent.

{
  "name": "remove_favorite_tld",
  "arguments": { "tld": "xyz" }
}

{
  "removed": "xyz",
  "total": 4,
  "limit": 200
}

Returns the domains bookmarked on the calling account, newest first. Use it to answer “what have I shortlisted so far?” without the person leaving the conversation, or to re-check a shortlist against search_domains before they buy.

{
  "name": "list_saved_domains",
  "arguments": {}
}

{
  "domains": [
    { "domain": "northbeam.dev", "savedAt": "2026-08-26T11:31:40Z" },
    { "domain": "sprouted.io",   "savedAt": "2026-08-24T09:02:11Z" }
  ],
  "total": 2,
  "limit": 500
}

Bookmarks a name on the calling account. It appears in the library on the site immediately and survives the conversation, which is the point: the shortlist outlives the chat window. Accepts a full domain or a bare label, the way the site saves a name being compared across extensions.

{
  "name": "save_domain",
  "arguments": { "domain": "northbeam.dev" }
}

{
  "saved": "northbeam.dev",
  "total": 3,
  "limit": 500
}

Removes one bookmarked name from the calling account. It is the counterpart to save_domain, for when a candidate is ruled out mid-conversation. Removing something that was never saved comes back as an error.

{
  "name": "remove_saved_domain",
  "arguments": { "domain": "sprigly.io" }
}

{
  "removed": "sprigly.io",
  "total": 2,
  "limit": 500
}

Returns the account's search history, most recent first. It is the cheapest way for an assistant to resume a naming session: what was already tried, and therefore what not to suggest again. This is what the person searched on the website: search_domains does not write here, so checking a name through an assistant never adds to it.

{
  "name": "get_recent_searches",
  "arguments": { "limit": 3 }
}

{
  "searches": [
    { "query": "northbeam", "searchedAt": "2026-08-26T11:32:02Z" },
    { "query": "harbor",    "searchedAt": "2026-08-26T11:31:58Z" },
    { "query": "ledger",    "searchedAt": "2026-08-25T16:20:44Z" }
  ],
  "count": 3
}

Returns the account's generation history, most recent first: the seed words, not the candidates. Useful for picking up where a naming session left off, or for noticing which direction someone keeps circling back to. Like the search history, it records the website only: generate_domains does not write here.

{
  "name": "get_recent_generations",
  "arguments": { "limit": 3 }
}

{
  "generations": [
    { "keyword": "orbit",  "generatedAt": "2026-08-26T11:31:51Z" },
    { "keyword": "bloom",  "generatedAt": "2026-08-26T10:14:03Z" },
    { "keyword": "harbor", "generatedAt": "2026-08-25T16:19:30Z" }
  ],
  "count": 3
}