Meta's Muse Sparked an "agent" Domain Spike: Analyzing Newly Registered Domains with DomainKits MCP

Disclosure: this post uses the domain name MCP from DomainKits.

Newly registered domains can be read as alternative data: a daily record of the names people decide to pay for. Several groups drive that record at once, including people building a product, people registering names to sell, and people holding names for later. This post reads one trend, "agent" in .com, against one reference keyword, "hub", then asks which group drove the change.

Key findings

• New .com names ending in "agent" spiked from 9 to 18 September 2026.
• The spike followed Meta's launch of its Muse agent on 8 September.

The setup: change one thing only

Raw registration counts move for many reasons, so this comparison holds everything fixed except the keyword:

Variable	Setting
Extension	.com
Characters	letters only
Position	the keyword ends the name (...agent.com, ...hub.com)
Tool	nrds, the 60-day index of newly registered domains
Window	50 days, 10 August to 28 September 2026
Periods	five periods of 10 days each
Keyword	"agent" or "hub"

Why "hub" as the reference

"Hub" is a reference keyword for this comparison, not a stand-in for the whole naming market. Two numbers from the server's keywords_trends tool (.com, letters only, last 28 days) support the choice:

• "hub" ranks 13th among the most registered .com keywords, with 14,005 registrations.
• 87% of those put "hub" at the end of the name, the position fixed in this comparison.

1. Newly registered domains by period

Period	Ending in "agent"	Ending in "hub"
10 to 19 Aug	740	4,119
20 to 29 Aug	560	4,087
30 Aug to 8 Sep	688	4,239
9 to 18 Sep	952	4,277
19 to 28 Sep	590	4,032

Figure 1: New letters-only .com names ending in "agent" or "hub" per 10-day period, indexed to 10 to 19 August.

"Hub" stays within 4% of where it started. "Agent" drops about 24% in late August, rises to 29% above its starting level from 9 to 18 September, then falls to 20% below it. keywords_trends, which counts "agent" anywhere in the name, puts its peak day on 12 September, inside the same peak period.

2. Search interest over the same 50 days

To compare registrations with public interest, we pulled daily Google Trends values for "ai agent" (worldwide, web search) for the same 50 days.

Period	"agent" names	Search interest, "ai agent" (period average)
10 to 19 Aug	740	44
20 to 29 Aug	560	34
30 Aug to 8 Sep	688	45
9 to 18 Sep	952	42
19 to 28 Sep	590	48

Figure 2: Daily Google search interest for "ai agent", 10 August to 28 September 2026.

In August, registrations and search interest moved together: both fell in late August and recovered in early September. From 9 to 18 September they moved apart: registrations rose while search interest slipped. Search interest reached its highest days on 20 and 21 September, and "agent" registrations fell in that period.

3. Why did this happen?

We found the same surge in Google Trends and in registrations. The trend changed, but what caused this change? Muse, Meta's new personal agent, launched on 8 September (announcement). Searches for "ai agent" rose on 9 and 10 September, and searches for "meta muse" jumped from 17 to 76 on 9 September. Names combining "muse" and "agent" numbered 39 from 31 July to 9 September, then 262 from 10 to 19 September.

Then who drove this change? Speculative domain sellers. The DomainKits nrds tool has a for-sale filter, has_sale: true. Adding it to the same query, with every other setting unchanged, gives the names in each period that are already listed for sale:

Period	"agent" names	Listed for sale	% of "agent" listed	% of "hub" listed
10 to 19 Aug	740	54	7.3%	1.1%
20 to 29 Aug	560	96	17.1%	1.0%
30 Aug to 8 Sep	688	129	18.8%	1.1%
9 to 18 Sep	952	392	41.2%	1.0%
19 to 28 Sep	590	84	14.2%	1.1%

Figure 3: New letters-only .com names ending in "agent" per 10-day period, split by sale status.

The listed share of "agent" rose from 18.8% to 41.2% in the peak period, while "hub" stayed between 1.0% and 1.1% in every period. Of the 264 extra "agent" names, 263 were listed for sale, and names not listed stayed level (559, then 560). The peak came from sellers.

Summary

Using the DomainKits MCP server and Google Trends, we saw a piece of tech news show up in domain registrations: after Meta launched Muse on 8 September, new names combining "muse" and "agent" surged. But almost all of these were registered by speculative sellers hoping for a quick flip. In reality, combining a corporate brand with a generic keyword carries a high risk of trademark infringement (and UDRP claims), meaning these "investors" moved fast, but likely into a legal dead end.

Run it yourself

In Claude, add DomainKits from the Claude directory: claude.ai/directory/domainkits.

In Claude Code, connect the server with:

claude mcp add --transport http domainkits https://api.domainkits.com/v1/mcp

Cursor and Gemini CLI use the same URL (setup notes).

Then ask:

Using DomainKits nrds, count new letters-only .com domains ending in "agent"
and ending in "hub" for the last 50 complete days, split into five 10-day
periods. Sort by registration date to find where each period starts. Repeat
with has_sale true to get the names listed for sale. Give the counts and
the listed share for each period.

Then compare the periods with Google Trends for the same dates. To test another keyword, change only the keyword and keep every other setting.

Data sources and method

• Registrations: the DomainKits MCP server, queried on 30 September 2026. The same data is available as a newly registered domains database.
• Totals, 10 August to 28 September: 3,530 "agent" names (755 listed for sale), 20,754 "hub" names (218 listed for sale).
• Search interest: Google Trends, "ai agent", worldwide web search, daily, scaled 0 to 100 against its highest day.
• Key nrds parameters:
  • tld: com, type: all_alpha: .com names made of letters only.
  • position: end: the keyword ends the name, so "agent" also matches words such as "reagent".
  • has_sale: true: names listed for sale; by default this covers the major domain marketplaces.
  • sort: reg_date_asc: oldest first, used to find where each 10-day period starts.
• keywords_trends (type: hot, scope: com): 28-day keyword ranking, used for the reference numbers and the peak day.