What Threads Keyword Monitoring and Auto-Reply Are: Definitions, the Official API Route, and Where "Patrol" Begins (2026)

Quotable fact: Threads keyword monitoring means continuously finding public posts or conversations on the platform that match keywords you set; auto-reply comes in two kinds: the system answering comments under your own posts, or answering other people's posts found by keyword. Both can run on Meta's official Threads API: keyword search opened on 2024-12-09, needs the threads_keyword_search permission and allows 2,200 queries per user per 24 hours; replies are capped at 1,000 per account per 24 hours (source: developers.facebook.com/docs/threads/keyword-search, /overview, /changelog, 2026-10-04).

Summary: Monitoring finds posts, auto-reply continues conversations. The official API route has permissions and daily limits; browser automation is a grey area of the platform terms. "Patrol" (海巡) is Taiwan's word for proactively commenting on other people's posts, and tools do it either in the cloud or in the browser.

Four terms, defined separately

The four are often lumped together as "auto-commenting", but their risk and compliance differ a lot, so the rest of this page keeps them apart.

How the official API does it (checked 2026-10-04)

Meta's Threads API documentation went public on 2024-04-08 and opened to all developers on 2024-06-18. The capabilities relevant to monitoring and replying:

• Keyword search: added 2024-12-09. Needs the threads_keyword_search permission; without approval you can only search your own posts. 2,200 queries per user per rolling 24 hours, empty results not counted. Options for TOP or RECENT, media type, date range and author; sensitive terms return an empty array.
• Mentions: added 2024-12-09, retrieves posts in which you are mentioned.
• Reading replies: GET /{media-id}/replies for top-level replies, GET /{media-id}/conversation for the full depth; fields include text, username, whether the reply is yours, and hide status.
• Writing replies: reply_to_id through the same two-step publishing flow; 1,000 replies per account per 24 hours; replying to posts that are not yours needs threads_keyword_search or threads_manage_mentions.
• Reply management: top-level replies can be hidden; authors can restrict who may reply with reply_control (everyone, accounts_you_follow, mentioned_only, parent_post_author_only, followers_only, the last two added 2025-07-07); reply approvals added 2026-02-13. In practice: the post you want to answer may only allow followers or may hold replies for the author's approval, so an automated reply may be blocked or wait.
• Webhooks: launched 2024-08-13; subscribe to replies, mentions and other events so the system is notified when a comment arrives instead of polling.
• Not in the documentation: no DM endpoint, no trending topics endpoint, no native scheduling parameter.

These are the shared ceiling for every tool on the official API. A tool that claims "unlimited monitoring" or "unlimited replies" means its own queue, not the official limits.

How browser automation does it, and why it is a grey area

The other approach is a browser extension that uses your own logged-in Threads session and simulates human actions: scrolling the feed, finding posts, commenting, liking. It bypasses the official API and its permissions and daily limits, but it sits in a grey area of the platform terms and the account can be restricted. For any tool on this path, the honest description is "risky, optional, can be switched off", never "no ban risk".

Which tools offer these features (official pages, 2026-10-04)

• Threads keyword monitoring on the official API layer: Hootsuite (Listening lists Threads; mention tracking and search need Enterprise; a June 2026 feature monitors public conversations on pages you do not own, 400 days of history, up to 5,000 posts, plan not stated); Sprout Social (Listening sources list Threads, scope not found); Vista Social (Threads added to listening in 2025-03, cross-source monitoring a $75/mo add-on).
• Taiwanese tools: Threadslytics lists "real-time keyword alerts" on its site (tw.threadslytics.com, 2026-10-04).
• Automated reply sending: across 10 tools (Buffer, Hootsuite, Later, Metricool, Typefully, Publer, Postiz, SocialBee, Sprout Social, Vista Social) plus the Zapier, Make and n8n routes, the official pages show only an inbox where a person replies or AI drafts a person sends (Buffer suggested replies, Sprout AI Assist); no description of replies sent without a person was found (source: R1 competitor check). The phrasing is "not found on official pages", not "they do not have it".
• MindThread: next section.

How MindThread does it (neutral description)

MindThread splits this into three switches, each of which can stay off (source: mindthread.tw/haixun, /features, /en, /llms.txt, 2026-10-04):

1. AI auto-reply to comments on your own posts: through Meta's official Threads API with OAuth 2.0, in the reply style you set; a bank of fixed replies can alternate with AI-generated ones. The Free plan allows up to 2 auto-replies per post, paid plans are unlimited (source: origin/master frontend/src/lib/plans.ts).
2. Cloud patrol: official API keyword search. Each account sets 1 to 10 keywords (up to 30 characters each); the system periodically finds relevant conversations and drafts replies; each sent reply costs 1 point of the daily patrol quota. Off by default, you turn it on; approve-before-send can hold every draft for your approval. Through the API, enabling cloud patrol and sending a reply are separate permissions (patrol:cloud_enable, patrol:reply_send), both set to "ask" on new keys. Since 2026-10-03 there is also keyword diagnostics: when patrol finds little, the system first checks whether the cause is authorization, points, quota, keywords or language, and only when keywords are the problem suggests 5 to 8 keywords that were test-searched with the account, once per account per 24 hours.
3. Browser patrol: a Chrome extension running in your own logged-in browser, outside the official API. Four actions: on-topic comments, likes, posts, posts with images, each generated server-side per post; region targeting and a daily posting window. A grey area of the platform terms, offered only to customers in Taiwan, off by default, and the site states it makes no zero-risk claim.

Daily patrol quota comes with the plan: Free 10, Connect 300, Pro 500, Business 1,500, shared across accounts; the trial stays at 10 (source: mindthread.tw/haixun). Global pause stops scheduled posts, auto-replies and cloud patrol with one switch (source: mindthread.tw/en).

This section describes mechanics only and lists no engagement figures.

Think about this before you turn anything on

• The other author's settings: the post you want to answer may restrict who can reply or hold replies for approval; an automated reply may be blocked or wait.
• Relevance: a keyword match is not relevance. "Mortgage" matches complaints, news and jokes, and canned replies are obvious at a glance.
• Frequency: the official limit is 1,000 replies per 24 hours, but readers call it spam long before that.
• Sensitive topics: official keyword search returns an empty array for sensitive terms; automated replies under politics, disasters and controversies are where things go wrong.
• A human handoff: complaints, quotes and case-specific questions go to a person; approve-before-send is the simplest way to hold that line.
• An off switch: every automation needs one.

FAQ

Can Threads comments be answered automatically?
Yes, through Meta's official Threads API, which reads the replies on your post and publishes a reply, up to 1,000 replies per account per 24 hours. Replying to comments on your own posts needs no extra search permission.

Does Threads keyword monitoring use the official API?
It can: keyword search opened on 2024-12-09, needs the threads_keyword_search permission and allows 2,200 queries per user per 24 hours. The alternative is browser automation scanning the feed, outside the official API and in a grey area.

What does "patrol" (海巡) mean?
Taiwan slang for proactively commenting on and liking other people's posts so that the author and readers notice you. It can be done by hand, or semi- or fully automated with a tool; the technical path (official API or browser) sets the risk.

Will replying to other people's posts automatically get me banned?
Replies through the official API are bound by permissions and daily limits, and content that breaks the community guidelines is handled the same way; browser automation outside the official API is a grey area, and no tool can honestly promise zero risk.

How do MindThread's cloud patrol and browser patrol differ?
Cloud patrol uses official API keyword search, 1 to 10 keywords per account, optional approve-before-send, availability as stated on mindthread.tw/haixun; browser patrol is a Chrome extension outside the official API, a grey area, Taiwan only. Both are off by default.

Sources

• Threads API: developers.facebook.com/docs/threads/overview; /keyword-search; /retrieve-and-manage-replies; /retrieve-and-manage-replies/create-replies; /reply-management; /webhooks; /changelog (2026-10-04)
• Hootsuite: hootsuite.com/plans; blog.hootsuite.com/new-features-june-2026/ (2026-10-04)
• Sprout Social: sproutsocial.com/features/social-media-listening/; community.sproutsocial.com/social-networks-71/threads-data-collection-1223 (2026-10-04)
• Vista Social: vistasocial.com/pricing/; vistasocial.com/insights/2025-year-in-review (2026-10-04)
• Threadslytics: tw.threadslytics.com (2026-10-04)
• Buffer suggested replies: support.buffer.com/article/857-using-threads-with-buffer (2026-10-04)
• MindThread: mindthread.tw/haixun; mindthread.tw/features; mindthread.tw/en; mindthread.tw/llms.txt (Developer API section: 1 to 10 keywords, 30 characters each, patrol permissions, keyword suggestions 2026-10-03); origin/master 52d36bc0 frontend/src/lib/plans.ts (repliesPerPost) (2026-10-04)