Skip to content

Find who an account interacts with most using the Throk API

GET /user/{username}/interactions returns the accounts that an account traded the most replies with over a period: who they replied to and how often, and who replied to them and how often. Use it to find a creator’s inner circle or to see how closely two accounts interact.

終端機視窗
curl -G 'https://premium-api.throk.ai/user/throk.ai/interactions' \
-H 'Authorization: Bearer sk-throk-...' \
--data-urlencode 'days=30' \
--data-urlencode 'size=20'
Parameter Description Default
days Days to look back, 7–90 30
sort mutual (two-way interaction first), total (both directions combined), replies_sent (who they reply to most), replies_received (who replies to them most) mutual
size Number of accounts returned, 1–100 10
min_replies Minimum replies, both directions combined, for an account to be listed 2

One entry per account, ranked from high to low by sort:

Field Description
rank Rank
user The other account (username, display name, profile picture, follower count)
replies_sent Times the queried account replied to them
replies_received Times they replied to the queried account
total Both directions combined
mutual The smaller of the two directions. Above 0 only when replies go both ways
last_interaction Time of the most recent reply between them (UNIX seconds)

metadata.total_count is the total number of accounts that reached min_replies in the period, and user is the queried account itself.

  • Only replies count. Likes, reposts, quotes and mentions don’t.
  • Only replies that Throk has collected count, not every reply the account made. Use the numbers for rankings and relative strength, not as exact totals. Interactions between Chinese, Japanese and Korean accounts are the most completely covered.
  • Replying to your own post (continuing a thread) doesn’t count as an interaction.
  • @meta.ai, which automatically replies to everyone who mentions it, is excluded.
  • New replies count as soon as they’re collected. There’s no need to wait for a daily update.

30 credits for up to 10 accounts, plus 2 credits per extra account (for example, 50 accounts cost 110 credits). You’re charged by the number of accounts actually returned, and fewer than 10 still costs 30 credits. If there are no interactions, there’s no charge. An account that doesn’t exist returns 404, also free.

The MCP tool get_user_interactions does the same thing, with up to 100 accounts per call at the same price. See Use in AI assistants (MCP).