Pass sort rank_change and window 30d on get_top_trends type Top Websites. Shorter windows return a monthly-feed note instead of climber rows.
Live data as of 2026-08-29
Top Websites climbers on Trends MCP come from get_top_trends with type Top Websites, sort rank_change, and window 30d. The August 29, 2026 pull still carried as_of_ts 2026-08-02T04:01:35 UTC because this feed updates on a monthly clock. 1d, 3d, 7d, and 14d each returned count 0 plus a note that points back to window 30d. With category computers-electronics-and-technology, the 30d list filled 10 climbers against baseline 2026-07-02T04:01:31 UTC. Category finance filled 18. Default limit is 25. Free tier access is 100 requests per month. For the current-rank board rather than climbers, see the Top Websites API.
On the August 29, 2026 Trends MCP pull, Top Websites sort rank_change kept as_of_ts 2026-08-02T04:01:35 UTC for every window, and every short window came back empty with an explicit retry.
| window | count | baseline_ts | note (abridged) |
|---|---|---|---|
| 1d | 0 | null | No baseline within 1 days of 2026-08-01. Feed updates roughly monthly. Retry with window 30d. |
| 3d | 0 | null | No baseline within 1 days of 2026-07-30. Retry with window 30d. |
| 7d | 0 | null | No baseline within 3 days of 2026-07-26. Retry with window 30d. |
| 14d | 0 | null | No baseline within 5 days of 2026-07-19. Retry with window 30d. |
| 30d | 10 (tech) / 18 (finance) | 2026-07-02T04:01:31 UTC | Climber objects with rank, keyword, prev_rank, rank_change, category_path. |
That note is the product contract. Daily feeds such as Google Trends rank_change can fill a 1d list against a same-day 02:01 snapshot. Top Websites does not. Copying a 1d GitHub or Wikipedia climber call onto this type yields count 0, not a shorter version of the 30d list.
The as_of_ts staying on August 2 while the pull date is August 29 is the same monthly cadence. A dashboard that treats Top Websites like a daily board will look stale for weeks even when the MCP call succeeds.
The MCP tool is get_top_trends. REST uses mode "get_top_trends". Type is the case-sensitive string Top Websites. category should be an official kebab-case name from the data-sources list, including computers-electronics-and-technology and finance.
{
"mode": "get_top_trends",
"type": "Top Websites",
"category": "computers-electronics-and-technology",
"sort": "rank_change",
"window": "30d",
"limit": 25,
"offset": 0
}
MCP clients pass the same fields without mode:
get_top_trends(type="Top Websites", category="computers-electronics-and-technology", sort="rank_change", window="30d", limit=25)
Default sort is rank, which returns [rank, domain, category] arrays. sort rank_change switches the rows to objects: rank, keyword, prev_rank, rank_change, and category_path. window is ignored unless sort is rank_change. Default window is 30d, which happens to be the only window that produced rows on this stamp. Setting 1d anyway is still useful as a probe: the note tells the client to retry 30d instead of guessing.
Omit category only when sampling the mixed board. Filtered tech and finance lists on this stamp stayed inside one category_path. The mixed 30d list did not.
Category computers-electronics-and-technology with window 30d, limit 25, offset 0 returned count 10 against baseline 2026-07-02T04:01:31 UTC. The current-rank board on the same stamp listed google.com at 1, facebook.com at 2, instagram.com at 3, x.com at 4, and brave.com at 25.
| rank | keyword | prev_rank | rank_change |
|---|---|---|---|
| 43 | vkvideo.ru | 46 | 3 |
| 44 | max.ru | 47 | 3 |
| 35 | threads.com | 37 | 2 |
| 4 | x.com | 5 | 1 |
| 7 | bing.com | 8 | 1 |
| 25 | brave.com | 26 | 1 |
| 27 | apple.com | 28 | 1 |
| 31 | music.youtube.com | 32 | 1 |
| 33 | telegram.org | 34 | 1 |
| 42 | aol.com | 43 | 1 |
The largest filtered tech moves were plus 3. x.com and bing.com each moved one slot inside the top 10. That is a sticky traffic board, not a daily hashtag list.
offset 10 with the same limit 25 then returned count 0 and the note "No items on the current 'Top Websites' board climbed versus the 30d baseline; everything held or fell." Those 10 rows were the full climber set. Pagination past the mover list is not a second page of climbers. It is a held-or-fell note. For limit and offset behavior across feeds, see get_top_trends pagination.
Category finance on the same 30d window returned count 18. The current-rank finance board opened paypal.com, tradingview.com, chase.com, investing.com, capitalone.com. Climbers included mufg.jp at rank 35 / prev_rank 43 / rank_change 8, rakuten-sec.co.jp at 14 / 18 / 4, discover.com at 42 / 46 / 4, schwab.com at 12 / 15 / 3, and investing.com at 4 / 6 / 2.
A 30d rank_change call with no category on this stamp filled count 25 at offset 0 and another count 25 at offset 25. Rows carried category_path values from construction, telecom, holidays, sports betting, and health, among others.
The first unfiltered page opened with ltk.com at rank 9, prev_rank 47, rank_change 38, category_path heavy-industry-and-engineering/construction-and-maintenance. directvgo.com sat at 11 / 47 / 36 under computers-electronics-and-technology/telecommunications. ytmp3.gg appeared twice: rank 15 / prev_rank 47 / rank_change 32 under community-and-society, and rank 7 / prev_rank 27 / rank_change 20 under community-and-society/faith-and-beliefs.
Those plus-30 jumps are not comparable to the filtered tech list, where the largest move was plus 3. An unfiltered Top Websites climber table is a mix of every category board. Agents that skip category will treat a construction-site domain as if it climbed the same list as google.com.
Official Top Websites category strings on Trends MCP include adult, arts-and-entertainment, business-and-consumer-services, community-and-society, computers-electronics-and-technology, e-commerce-and-shopping, finance, food-and-drink, gambling, games, health, heavy-industry-and-engineering, hobbies-and-leisure, home-and-garden, jobs-and-career, law-and-government, lifestyle, news-and-media, pets-and-animals, reference-materials, science-and-education, sports, travel-and-tourism, and vehicles. The string is case-sensitive in the same way as other typed feeds.
get_growth on google search is a separate request from the board. On August 29, 2026 the follow-ups used the latest Google Search points the series had, which is not the August 2 Top Websites stamp.
investing.com sat at finance rank 4 after a 30d climb from 6. Google Search for the same string sat at index 19.0 on 2026-08-22 versus 34.0 on 2026-07-25, growth -44.12% over 30D, then 19.0 versus 60.0 over 3M (-68.33%) and 19.0 versus 99.0 over 12M (-80.81%). The series metadata reported 260 points. Traffic rank and search demand moved in opposite directions on that name.
Charles Schwab (the brand string used for schwab.com) sat at finance rank 12 after a climb from 15. Google Search returned recent_value 55.0 on 2026-08-15 versus 52.0 on 2026-07-18, growth +5.77% over 30D, with estimated volume 3,200,000 versus 3,030,000. The 3M window was 55.0 versus 71.0, growth -22.54%. One 30d climb on Top Websites does not imply a 3M search rise.
threads.com climbed 37 to 35 on the tech board. Google Search for threads sat at 10.0 versus 11.0 over 30D (-9.09%), estimated volume 3,640,000 versus 4,000,000, and 10.0 versus 10.0 over 12M (flat). bing climbed 8 to 7 on the board while Google Search sat at 17.0 versus 18.0 over 30D (-5.56%) and 17.0 versus 22.0 over 12M (-22.73%). Keyword x.com returned data_unavailable on google search. X Twitter sat at 34.0 versus 44.0 over 30D (-22.73%); twitter sat at 28.0 versus 29.0 (-3.45%). Domain rank and search keyword are not interchangeable labels.
Volume growth on google search was omitted in several of these payloads because volume is derived from the trend value for that source. Quote the two index values. Do not invent a second independent volume series. Brand search follow-ups for this feed also appear on Top Websites vs Google Trends.
FAQ