REST API

Posts

Publish, read, retry and delete.

GET/v1/posts

Recent publishes, newest first. Replies are not listed as posts.

Paging

NameTypeDescription
limitquery1-100, defaults to 20.
beforequeryA publish id from a previous page's nextBefore. Returns the ones older than it.
includeDeletedquerytrue brings back posts you deleted from the channel. They are hidden by default. Our record and its metrics stay either way.
statusqueryNarrow to one or more statuses, comma separated: scheduled, draft, publishing, published, partial, failed or cancelled. Omit for everything, scheduled and draft included.
sincequeryYYYY-MM-DD or ISO 8601 with a four-digit year. Posts dated at or after this. A scheduled post is dated by scheduledAt, a sent one by when it was created. A value that is not a real date, or a year past 9999, is refused with invalid_params.
untilqueryYYYY-MM-DD or ISO 8601 with a four-digit year. Posts dated at or before this; a date means the whole day. A value that is not a real date, or a year past 9999, is refused with invalid_params.
sortqueryscheduled orders by the post's date ascending and drops paging. Omit for newest first.
hasMoreresponseThe list was cut short.
nextBeforeresponsePass it back as before. Absent when there is nothing more.

POST/v1/posts

Publish to one or more connected accounts.

NameTypeDescription
contentRequiredstringPost text. Emoji count as UTF-8 bytes.
accountIdsRequiredstring[]Ids from GET /v1/accounts.
mediaIdsstring[]Media already confirmed with complete. On Naver Blog you may leave out ids the body already points at with media:<id>; we read those from the text and attach them.
optionsobjectPer-channel settings, keyed by the channel id. The shape is in GET /v1/platforms under options and on each channel page. Some channels cannot publish without it β€” YouTube needs options.youtube.title and Naver Blog needs options.naver_blog.title plus a category. Every channel takes options.<channel>.content, a different text for that channel alone.
scheduledAtstringISO 8601 with a timezone offset, 10 minutes to a year out. We keep the post as scheduled and send it at that time, on every channel.
draftbooleantrue saves the post without sending. It can have no accountIds yet. Send it later with POST /v1/posts/{id}/publish.
batchIdstringOptional tag, up to 40 characters, to group posts made together (for example the same text sent to two workspaces). Comes back on every post.
replyTostringThe post or reply to answer: its id on the channel, our publish id (post_…), or a link to the post. A publish id or a link becomes that post's id on the replying account's channel, so send one account per request. A Threads, Instagram, Facebook or Bluesky link works only for a post we have on record.
threadItemsobject[]Split a long post into a chain: [{ content, mediaIds? }]. The first item is the root, the rest become replies under it in order. Each item obeys the character limit on its own. Exclusive with content. Only Threads, Bluesky and X take a chain; the composer builds one with the item editor. If a later item fails the ones already up stay up, and the response carries thread.resumeFrom. If an item fails, the later items on that account close as failed with errorCode thread_broken, and retry_post continues from the item that failed. Posts that have later items carry them back as threadItems, and PATCH takes the same field while the post is scheduled or a draft.
topicTagstringOne topic to tag the post with. Threads takes exactly one and rejects periods and ampersands. On a thread it applies to the first piece only. This top-level field is an alias: if a channel in the call does not take topics, the whole call is refused. To tag only the Threads part, use options.threads.topicTag.
Idempotency-KeyheaderSend the same key when you retry and we return the first answer instead of publishing twice. Reusing a key with a different body is a 422. Kept for 24 hours.
waitbooleanHold the response until the post is really out. Text waits up to 10 seconds, media up to 75. Past that you get the usual 202 with a next hint, and the post keeps going. Defaults to false.
curl -X POST https://api.uplika.com/v1/posts \
  -H "authorization: Bearer $UPLIKA_KEY" \
  -H "content-type: application/json" \
  -d '{
    "content": "Now with images",
    "accountIds": ["acc_01H..."],
    "mediaIds": ["med_01H..."]
  }'

GET/v1/posts/{id}

One publish with the status of each target.

Target status

NameTypeDescription
publishingstatusIn flight at the channel.
publishedstatusCarries externalId and url. If the account was disconnected afterwards, the target stays published with connectionId null and accountHandle set to that account's handle until the same account is reconnected.
partialstatusSome channels went out and some did not. Retry sends only the ones that failed. There is no 207 β€” publishing is 202 plus polling, so nothing is held open.
cancelledstatusDeleted from the channel. Our record and its metrics stay. Hidden from GET /v1/posts unless you pass includeDeleted=true.
failedstatusCarries error.
scheduledstatusHeld until scheduledAt. Nothing is on any channel yet. Editable.
draftstatusSaved without a time. May have no targets. Editable.

A publish has one row per channel. publishing is in flight, published carries the channel's own id and URL, failed carries a reason.

What an X post cost

On X targets only, the target row carries billing: { operation, mills, state, balanceMills }. operation is post, post_url for a post containing a link, or reply for a reply, which includes each piece of a chain after the first. mills is what that post cost from X credits in thousandths of a US dollar, attached files and alt text included. state is reserved while it is going out, confirmed once X took it, or released when it was handed back. balanceMills is the person's X credit balance now. It is on single-post responses (GET /v1/posts/{id}, and the post in GET /v1/posts/{id}/thread) and on publish responses. List endpoints do not carry it. When the balance cannot cover a post, publishing answers 402 x_credits_insufficient with topUpUrl and nothing goes out.

What can go in {id}?

Any way you have of naming the post: the publish id (post_…), the post's own id on the channel, its shortcode, or a link to it. Put a link through encodeURIComponent so it stays one path segment. postId on POST /v1/engagement/like and on the /v1/replies/{replyId} calls takes the same values. A post written in the channel's app is found on that channel the first time and kept as an imported record, so replies, insights, edits and deletes work on it; the next call with the same link reads our record and does not call the channel. Finding a shortcode reads the account's recent posts (up to 500), so an older one comes back post_not_in_recent with scanned, and its id on the channel works instead. A post on a channel this workspace has not connected is post_not_owned. A Naver post on someone else's blog is naver_wrong_blog: reply to it or like it with ownerHandle and externalId on POST /v1/engagement/reply or /v1/engagement/like. A YouTube link on a day uplika's YouTube allowance is used up answers 429 youtube_quota_exhausted. replyTo on POST /v1/posts takes a publish id or a link too, and we send that post's id on the replying account's channel. A Threads, Instagram, Facebook or Bluesky link works only for a post we have on record. A link or a channel's own id names one channel's copy, while DELETE and an edit of a published post (PATCH) act on every channel the post went out on: when the post is also on other channels they answer 409 post_spans_channels with postId and channels and change nothing. Call again with that postId to act on every channel.

Scheduling and drafts

Pass scheduledAt and we keep the post as scheduled and send it at that time, on every channel, from our side. Pass draft: true and nothing goes out until someone calls publish or the dashboard does. Both come back with status scheduled or draft, can be changed with PATCH, sent early with /publish, and removed with DELETE. scheduledAt needs a timezone offset and is 10 minutes to a year out. Some channels also have scheduling of their own (options.youtube.publishAt, options.facebook.scheduledPublishTime, options.naver_blog.scheduledAt); those still work over the API but the dashboard hides them, because two clocks on one post is a contradiction waiting to happen.

Different text per channel

Every channel takes options.<channel>.content. When set it replaces the shared content for that channel only, and the length and media checks use it. That is how a 500-character Threads post and a Naver Blog Markdown body fit in one call. GET /v1/platforms lists what else each channel takes, under options.

PATCH/v1/posts/{id}

Change a scheduled or draft post before it goes out. Fields you leave out keep their value, except that a post going only to Naver Blog takes its media list from the new body's media: references when you leave mediaIds out; options replaces the whole object; threadItems replaces a chain, content alone edits its first item. Editing a post that is already live is asynchronous: wait: true holds the response for the result (up to 75 seconds), and past that it answers 202 with next telling you to read the post. A Naver draft save that went quiet is looked up in Naver's draft list and recovered when it is there. A post published with publish_naver_draft is edited like a post written in Naver's editor: the first update returns its current source (409 naver_source_required) and the rewrite goes out with sourceVersion, so the photos and formatting Naver has stay. A Naver post published through uplika takes the same path once its uploaded photo files have been cleaned up (about a day after it went out): changing only options goes out directly, and changing the body returns the current source first.

NameTypeDescription
contentstringNew post text.
accountIdsstring[]New target accounts. Replaces the current set.
mediaIdsstring[]New media, in order. Replaces the current set. On a post going only to Naver Blog, leave it out when the body places media with media: references: those references become the list and media the new body does not place is removed (the response note says how many).
optionsobjectPer-channel settings. Replaces the whole object.
scheduledAtstringNew send time, same rules as on create. null turns the post into a draft.
draftbooleantrue turns a scheduled post, or a Naver Blog post still waiting for the browser extension, back into a draft.

POST/v1/posts/{id}/publish

Send a scheduled or draft post right now. Same function the scheduler runs, so it cannot go out twice.

NameTypeDescription
waitbooleanHold the response until the post is really out, same as on create.

POST/v1/posts/{id}/retry

Retry the targets that failed. Targets that already went out are skipped. A Naver target that stopped right after Save as draft (naver_draft_unknown) a Naver reply that did not confirm in time (naver_reply_unknown), a WordPress post the site never confirmed (wordpress_publish_unconfirmed), a Naver Clip that was posted but not confirmed in the clip list (clip_unconfirmed, check Clip Creators > Content first), an X post that was sent and then got no usable answer (x_publish_unknown, check the X account first), a TikTok post that was sent and then lost its connection before TikTok answered (tiktok_publish_unknown, check the TikTok profile first), a post that is live but held by an imported copy of it (published_as_imported_copy, use that copy), or a target that stopped before the channel answered (publish_outcome_unknown, check that channel first; if it is there, send the other channels as a new post) answers 409 until the body carries {"force": true}. If another post with the same content and the same files is already on the same account (a file uploaded again counts as the same file), it answers 409 duplicate_post with that post's id and retries nothing; {"force": true} sends it anyway.

DELETE/v1/posts/{id}

Delete the post from every channel it went out on and mark our record deleted.

On a scheduled or draft post nothing is on any channel yet, so DELETE simply removes our record. On a post that went out it deletes from every channel it went out on first and keeps our record marked deleted. If the post was named by a link or a channel's own id and it is also on other channels, DELETE answers 409 post_spans_channels with postId and channels and calls no channel; the same call with that postId deletes it everywhere. If a channel's trash is turned off, so the post is gone for good, the response adds note (the sentence to pass on) and permanentlyDeleted (the channel names). A Naver draft is removed by the browser extension when its turn comes in that browser; if that has not finished within 60 seconds, the response adds note saying so and the extension finishes it within 30 minutes.

Deleting a post on X is charged X's delete price from the person's X credits. If the balance is too low, DELETE answers 402 x_credits_insufficient with topUpUrl. The balance is checked before any channel is called, so nothing is deleted; if another call spent it in the meantime, the message names the channels already deleted.

Reading what is already on the channel

Our history only has what went out through us. These four read the channel itself and pick which channels to publish to. They are the routes behind list_channel_posts, get_quota, open_post and select_channels, so an agent gets the same answers over REST.

GET/v1/posts/{id}/thread

One post with its replies and metrics in a single call.

NameTypeDescription
accountIdqueryWhich channel's copy to read, on a post that went to more than one channel. Omit and the answer says which copy it read.

GET/v1/channels/posts

What is on the channel right now, including posts written in the app.

Query

NameTypeDescription
accountIdqueryLimit to one connected account. Omit to cover every channel.
limitquery1-100, defaults to 25.

GET/v1/channels/quota

How much of the 24 hour allowance the channel says is used. When an X account is in scope the response also carries xCredits: { balanceMills, topUpUrl }, the person's prepaid X credit balance in thousandths of a US dollar (0 when nothing was ever bought) and where to top it up.

GET/v1/channels/bridge

Whether the Naver Blog bridge (browser extension) is online, and how to wake Chrome if it is not.

GET/v1/channels/publish-options

What a channel needs to know before you publish to it. TikTok only: the creator nickname the post goes out as, the privacy levels this account may use right now, whether it can post at all, and its video length limit. Other channels return not_supported.

NameTypeDescription
accountIdqueryRequired. These values are per connected account and change when the person edits their channel settings, so do not cache them.

GET/v1/channels/instagram/audio

Music and original sounds one Instagram account can put on a reel, for options.instagram.audio. Instagram only, and only for accounts connected via Facebook. Without query it lists what is trending for that account; with query, music comes only from Meta's royalty-free Sound Collection, so an artist or song name will not find that song. There is no sort or date option, no count of reels using a sound and no creation date, matching is strict about spelling and spacing, and licensed songs in the trending list come with no audio file.

NameTypeDescription
accountIdqueryRequired. A connected Instagram account. Trending audio differs per account, so use ids from the same account's results.
typequerymusic (default) or original_sound.
queryqueryOptional, up to 100 characters. Without it you get this account's trending audio. With type music it searches royalty-free music from Meta's Sound Collection only. Matching is strict about spelling and spacing.
limitquery1-50, defaults to 25.
cursorquerynextCursor from the previous page, unchanged.

POST/v1/accounts/select

Pick which connected channels to publish to.

POST/v1/accounts/:id/refresh

Re-read a channel's metadata, such as Naver Blog categories or Pinterest boards. Waits up to a minute for the browser extension.

NameTypeDescription
scopestring | string[]Omit to list candidates. "all" for every active channel, or an array mixing platform names, handles and account ids.

PUT/v1/accounts/:id/pinterest/default-board

Set the board a Pinterest connection pins to when a post leaves out options.pinterest.boardId. The board must be in the connection's board list; call refresh first for a board made after connecting.

NameTypeDescription
boardIdstringRequired. A board id from the account's pinterest.boards in GET /v1/accounts. Answers with the account, pinterest.defaultBoardId updated.

GET/v1/naver/blog/diagnose

Read a Naver blog's public RSS feed (its latest posts, at most 50) and say whether the titles are written for search: the share of titles carrying a search intent word (price, how to, review), how many start with a date or episode label, posts per month, the categories, and the words the blog repeats in titles (candidates to research, not proven keywords). The response carries the date range it read, so a quiet blog's 50 posts are not mistaken for everything. Cached 24 hours across users; a cache miss spends one of 20 diagnoses per person per day. No Naver API keys are needed.

NameTypeDescription
blogRequiredstringA Naver blog id (2-40 letters, digits, - or _) or any link to the blog: blog.naver.com/<id>, m.blog.naver.com/<id>/<logNo>, PostView.naver?blogId=<id>.

POST/v1/naver/keywords/judge

Measure Naver search keywords and pick a set for one post. Expands nothing; send the keywords you have. Returns monthly searches (Search Ad), blog document count and posts per month (API HUB), a verdict per keyword, and sets of one main keyword plus two to five sub keywords with the same intent. Results are cached for seven days across users; 60 calls and 600 measured keywords per user per day.

NameTypeDescription
keywordsstring[]Required. 1-300 keywords. Spaces are fine; the numbers come back per keyword as you sent it.
savebooleanDefault false. true stores this research as a report (the keywords, the first set's main and subs, every set with its prompt) and returns reportId. The same person sending the same keyword set within ten minutes gets the existing reportId back with no new row and no judge call spent.
seedTextstringWhat the person typed, kept with the saved report as its label. Defaults to the first five keywords.
cacheOnlybooleanDefault false. true never calls Naver: keywords without a fresh cached value come back in unmeasured with no numbers. Use it after measuring in chunks, so the judge call is instant and spends nothing upstream.
budgetMsnumberTime budget for new measurements, default and maximum 60000. When it runs out, no new upstream call starts; the response says partial: true and lists unmeasured. Call again and the cached part is skipped.

POST/v1/naver/keywords/measure

Same measurement without the verdict and sets. withDocs:false skips document counts.

NameTypeDescription
keywordsstring[]Required. 1-300 keywords. At most 40 of them are measured fresh in one call; the rest come back in unmeasured and are picked up by the next call.
withDocsbooleanDefault true. false measures search volume only.
budgetMsnumberTime budget for new measurements, default and maximum 60000. Past it the call stops starting upstream requests and returns partial: true with unmeasured.

POST/v1/naver/keywords/expand

Expand seed keywords one level through Naver autocomplete, for agents and API keys (the dashboard does this in your browser). When the person's uplika Chrome extension is on, it looks the seeds up in the background from their own connection; otherwise our server does. Returns every word found with which seed produced it, and via says which path answered. When neither can run this answers 503 naver_autocomplete_unavailable, and you should pass your own keyword list to judge instead.

NameTypeDescription
seedsstring[]Required. 1-10 seed keywords. Each is sent to Naver autocomplete once (one level); the words that come back are candidates to measure with judge, not proven keywords. Cached seven days across users. No daily quota: when the person's uplika Chrome extension answers there is no wait, and when our server answers, one call per person every 60 seconds (naver_autocomplete_busy with retryAfterSeconds).

What comes back

NameTypeDescription
found{ keyword, from }[]Every word, the seeds first. from is seed, or L1:<seed> for a word that seed produced.
errors{ query, error }[]Seeds that could not be looked up, with why.
cachednumberHow many seeds came from the seven-day cache.
via{ cache, extension, server }How many seeds each path answered: the shared cache, the person's Chrome extension (in the background, from their own connection), or our server.

GET/v1/naver/keywords/reports

List the keyword research you saved (judge with save: true), newest first: when, what you typed, the first set's main and sub keywords, and every set with its ready-to-paste prompt. Reports belong to the person, so an API key limited to some workspaces gets 403.

NameTypeDescription
limitnumberDefault 50, newest first.

GET/v1/naver/keywords/history

The measurement history of one keyword: one point per day it was actually measured (cache hits add nothing), newest first, with search volume, document count and posts per month. An empty list is 200 with data: [].

NameTypeDescription
keywordRequiredstringThe keyword, as you would send it to measure. A query parameter, since keywords are usually Korean.
limitnumberDefault 100 days, newest first.
tznumberThe viewer's time zone offset in minutes, with the sign of JavaScript getTimezoneOffset (Korea is -540). Measurements fold into that zone's days. Default 0 (UTC).
NameTypeDescription
qRequiredstringThe search words, as a viewer would type them.
orderstringviewCount (default), date or relevance.
periodstring7d, 1m, 3m, 6m or 1y. Omit for no limit.
formatstringall (default), shorts or long. 60 seconds or less counts as a Short.
maxnumber1-50, default 25.
topicstringThe video the person wants to make, one line. Goes into prompt and is saved with the report.

GET/v1/research/youtube/reports

The person's saved YouTube searches, newest first: when, the search and its filters, the topic, how many videos and whether it came from cache. Results are kept for up to 30 days, then deleted. History belongs to the person, so an API key that does not cover all of the person's workspaces gets 403 forbidden_scope.

NameTypeDescription
limitnumberDefault 50, newest first.

GET/v1/research/youtube/reports/{id}

One saved search with the videos as they were and the prompt built from them. History belongs to the person, so an API key that does not cover all of the person's workspaces gets 403 forbidden_scope.

POST/v1/google/keywords/research

Google search volume for up to 10 keywords in a country the person chooses: average monthly searches, the last 12 months, ad competition and top-of-page bids, plus related keywords with the same numbers. Bids come in bidCurrency, the currency of uplika's Google Ads account, the same for every country and not converted. Search counts are Google's rounded figures. A keyword with no data from Google is listed in noData, which is not zero. Put every keyword in one search; a comma-separated string works too. The same search is cached 7 days. There is no per-person daily limit. Beta: the person can turn it off for API keys.

NameTypeDescription
keywordsRequiredstring[]1 to 10 keywords, as the person wrote them. A comma-separated string or an item with commas is split into keywords.
countryRequiredstringTwo-letter country code such as KR or US. Required, with no default. Ask the person which country they mean.
languagestringLanguage code such as ko or en. Leave it out to use the country's main language.
maxnumberHow many related keywords to return, biggest first. 1-200, default 50.
NameTypeDescription
qstringKeywords, as a viewer would search them, up to 100 characters. Pass q or advertiserId, not both.
advertiserIdstringA Facebook page id from /v1/research/ads/advertisers. Lists that advertiser's ads instead of searching words.
countrystringTwo-letter code of the country the ads were shown in. Default KR.
statusstringactive (default), inactive or all.
mediastringall (default), image or video.
sortstringimpressions (default) or relevance. There is no newest-first sort; narrow with from and to instead.
fromstringOnly ads that started on or after this date, YYYY-MM-DD.
tostringOnly ads that started on or before this date, YYYY-MM-DD.
exactbooleanDefault false. true keeps only ads with the exact phrase.
koreanOnlybooleanDefault true. Drops ads without Korean text, since country KR means ads shown in Korea and foreign ones are mixed in. received and kept say how many were dropped.
cursorstringThe cursor from the previous page.

GET/v1/research/ads/advertisers

Facebook pages that match an advertiser name in Meta's Ad Library, verified first and then by likes, with pageId, category, likes and the Instagram handle. Several pages often share a name, so let the person pick, then pass that pageId to the search as advertiserId. Spends one daily ad search; the same name within 24 hours is cached and free.

NameTypeDescription
qRequiredstringA brand or page name, up to 100 characters.

GET/v1/research/ads/{adId}/transcript

What is said in a video ad, by adId from a search result with hasVideo true. available is false when the ad has no video, no speech, runs longer than two minutes, or the data provider could not read the video. Saved for 30 days; only a new transcript spends one daily ad search.

GET/v1/research/ads/reports

The person's saved ad searches, newest first: when, keyword or advertiser, country, how many ads and whether it came from cache. The ads themselves are kept for 7 days (thumbnail links expire sooner) and the list for 30 days. History belongs to the person, so an API key that does not cover all of the person's workspaces gets 403 forbidden_scope.

NameTypeDescription
limitnumberDefault 50, newest first.

GET/v1/research/ads/reports/{id}

One saved ad search with the ads as they were. After 7 days resultsExpired is true and ads is null, so search again. History belongs to the person, so an API key that does not cover all of the person's workspaces gets 403 forbidden_scope.

NameTypeDescription
domainstringA website host such as oliveyoung.co.kr, up to 253 characters. A pasted address is reduced to its host. Pass domain or advertiserId, not both.
advertiserIdstringA Google advertiser id (AR followed by digits) from /v1/research/google-ads/advertisers or from an ad's advertiser.id.
regionstringTwo-letter code of the country the ads were shown in. Default KR.
platformstringall (default), search, youtube, shopping, maps or play.
formatstringall (default), text, image or video.
fromstringOnly ads shown at any time on or after this date, YYYY-MM-DD.
tostringOnly ads shown at any time on or before this date, YYYY-MM-DD.
cursorstringThe cursor from the previous page.

GET/v1/research/google-ads/advertisers

Advertisers in Google's Ads Transparency Center that match a brand or company name, largest adsEstimate first, plus matching websites in domains. Google keeps one advertiser per registered company or account, so one brand often has several: let the person pick, then pass that advertiserId to the search. A brand name that differs from the company name can find nothing; try the company's legal name or its website. Spends one daily ad-analysis lookup; the same name within 24 hours is cached and free.

NameTypeDescription
qRequiredstringA brand or company name, up to 100 characters.
regionstringTwo-letter country code to search advertisers in. Default KR.

GET/v1/research/google-ads/reports

The person's saved Google ad lookups, newest first: when, website or advertiser, country, how many ads and whether it came from cache. Meta ad searches are not in this list. The ads are kept for 7 days and the list for 30 days. History belongs to the person, so an API key that does not cover all of the person's workspaces gets 403 forbidden_scope.

NameTypeDescription
limitnumberDefault 50, newest first.

GET/v1/research/google-ads/reports/{id}

One saved Google ad lookup with the ads as they were. After 7 days resultsExpired is true and ads is null, so look it up again. History belongs to the person, so an API key that does not cover all of the person's workspaces gets 403 forbidden_scope.

POST/v1/naver/visibility/checks

Checks up to 10 keywords on Naver's integrated search through the person's uplika Chrome extension, which reads the public result page in their own browser about 3 seconds apart, without cookies or a window. Per keyword: whether an AI briefing answers it and in which form, its body and image sources with ranks and how many answer paragraphs cite each, where blog posts sit in the results, and the related questions. 202 with runId, or 200 with results when waitMs is given. A daily allowance of 200 pages per person; 503 bridge_offline when the extension is not on. A 200 answer also carries citationRules: sentence rules uplika measured on posts the AI briefing quoted (rules) and weaker suggestions with their grade, evidence and caveat (suggestions). They are uplika's measurements, not Naver's criteria. With blogTab: true it also reads page 1 of the blog tab sorted by relevance for each keyword and returns blogTab and verdict.blogTabRank (the blog's place among the blog posts there), spending two pages per keyword. Checks belong to the person, so an API key that does not cover all of the person's workspaces gets 403 forbidden_scope.

NameTypeDescription
keywordsRequiredstring[]1-10 keywords, as a reader would search them.
blogstringA Naver blog id or any link to the blog. Its posts are marked mine: true and each result carries verdict (bodyRank, imageRank, citedParagraphs, organicRank, organicArea, cited, exposed).
freshbooleanDefault false. A keyword this person read in the last 30 minutes comes back from the saved result; true reads it again and spends the daily allowance.
blogTabbooleanDefault false. true also reads page 1 of the blog tab sorted by relevance for each keyword: each result carries blogTab (read, sort, size, top 5 posts, reason, requiredVersion when the extension is too old) and, with blog, verdict.blogTabRank. Two search pages per keyword from the daily allowance. A saved result from the last 30 minutes is reused only if its blog tab was read.
waitMsnumberHold the response up to this many milliseconds (at most 60000) and answer 200 with the results so far; partial: true with pending lists what the extension is still reading in the background. Without it the answer is 202 with runId.

GET/v1/naver/visibility/runs/{runId}

One check: status (running, done, partial, failed), how many keywords came back, results in the order asked, and pending for the ones still being read. Once it is done, partial or failed it also carries citationRules, as on the check.

GET/v1/naver/visibility/runs

This person's recent checks, newest first: kind, keyword count, how many came back, and for a blog, how many were cited or shown.

NameTypeDescription
limitnumberDefault 20, at most 50, newest first.

POST/v1/naver/visibility/scan

Starts a blog's report card: reads its latest posts from the public RSS feed, picks each post's main keyword, and checks them in the background. 202 with runId and etaSeconds; a report card for the same blog from the last 6 hours comes back as 200 with reused: true.

NameTypeDescription
blogRequiredstringA Naver blog id or any link to the blog.
limitnumberHow many recent posts, 1-30, default 20.

GET/v1/naver/visibility/blog

A blog's report card from what was recorded: cards (AI citation rate, search exposure rate, keywords with an AI briefing, cited posts), axes (five scores out of 10 on uplika's own scale, each with its raw value and threshold; not a Naver score), firstThing, posts with their main keyword and latest verdict, categories, trend and events.

NameTypeDescription
blogRequiredstringA Naver blog id or any link to the blog.
tznumberThe viewer's time zone offset in minutes, with the sign of JavaScript getTimezoneOffset (Korea is -540). The trend groups checks into that zone's days. Default 0 (UTC).

GET/v1/naver/visibility/posts/{blogId}/{logNo}

One post's record: the post being watched (main keyword and where it came from), each check of that keyword with the post's verdict, and the events (started or stopped being cited or shown).

PATCH/v1/naver/visibility/posts/{blogId}/{logNo}

Sets a post's main keyword, the one its checks search for. Answers the post's record.

NameTypeDescription
mainKeywordRequiredstring1-100 characters. Used from the next check on; the post's old verdict is cleared.
titlestringThe post title, used only when the post is not tracked yet (up to 300 characters).

GET/v1/naver/visibility/keywords

Every check of one keyword by this person, newest first, each with the AI briefing's sources and the search results' blog posts.

NameTypeDescription
keywordRequiredstringThe keyword, up to 100 characters. A query parameter, since keywords are usually Korean.
blogstringMark that blog's posts and add a verdict to each check.
limitnumberDefault 30, at most 100, newest first.

POST/v1/naver/seo/check

Run the Naver Blog placement checks on a Markdown body before publishing: section titles match the sub keywords one to one, the main keyword appears 3-5 times, 12-15 tags all present in the body, no product names in the first three paragraphs, every attached image placed with a caption of ten or more characters, and a body length floor. Nothing is blocked; the report says what is off and why. The response also carries `pentagon` (the same five gate measurements as 0-10 scores: keyword, structure, length, photos, tags) and `citation` (counts of the sentence shapes AI briefings were seen quoting: definitions, how-to steps, specifics in parentheses, the first definition near the top, link cards). Both are information only.

NameTypeDescription
contentstringRequired. The Markdown body, up to 60,000 characters.
titlestringThe post title.
mainstringThe main keyword. Without it the keyword checks cannot pass.
subsstring[]Sub keywords, one per section title.
tagsstring[]The tags you will send in options.naver_blog.tags.
fixedTagsstring[]Tags that need not appear in the body (brand tags).
productsstring[]Product and brand names that must stay out of the first three paragraphs.
mediaIdsstring[]The media ids attached to the post, to check that each is placed.

POST/v1/naver/layout

See how a body will be laid out before publishing. The same pass publish runs by default (layout: template), returned without sending anything; forms: "all" lays it out in every preset so a person can pick one.

NameTypeDescription
contentRequiredstringThe markdown body to lay out.
mediaIdsstring[]Attached media, in order. Images the body does not reference are placed by the form.
formstringOne preset id (default photo-story). Ignored when forms is given.
forms"all" | string[]Lay the same body out in several presets at once; the response carries layouts[] so a person can compare and pick.

GET/v1/naver/forms

The seven post forms we ship. A form is the shape of a post: what opens a section, whether a quotation is a heading or a quoted passage, how often a sentence is highlighted, how many images per section, whether images get captions. Each one was measured from a real published post. Pass an id to options when generating; the form only shapes what we ask the model for, and never touches publishing.

GET/v1/naver/drafts

The drafts in the blog's draft box: posts saved with draftOnly and anything saved by hand in the Naver editor. Extension 0.7.2 or later.

POST/v1/naver/drafts/{logNo}/publish

Publishes a draft from the draft box exactly as it is in Naver: the extension loads it in the editor and presses publish, so hand edits are kept. Same response shape as POST /v1/posts; if uplika made that draft, that post record flips to published. To change the post after it is published, call update_post on it; the current source comes back first.

NameTypeDescription
logNoRequiredstringThe draft's id from GET /v1/naver/drafts.
accountIdstringRequired only when the workspace has more than one Naver Blog account.
title / categoryId / category / tags / openTypestringOverride the draft's own values. Left out, the draft goes out as it is.
waitbooleanHold until the extension finishes, up to 75 seconds like a post with media on POST /v1/posts. Past that you get the usual 202 with a next hint, and the draft keeps going out.

GET/v1/naver/grammar

What the Naver Blog body understands: directives, block and inline markup, attributes, highlights, media references and the limits. This is the same guide the describe_grammar tool returns, and the values come from one source, so it never drifts from what the editor actually accepts. Pass topic to read one section.

NameTypeDescription
topicstringOne section instead of the whole guide. The 400 lists the ids.

POST/v1/naver/forms/extract

Read a public Naver Blog post and hand back its form. We take the structure and the formatting, never the words or the photos. Use it for "write it like this post". We keep what we read, so AI generation uses that saved form and never goes back to Naver. Call it again to pick up changes in the reference.

NameTypeDescription
urlstringRequired. A public Naver Blog post, like https://blog.naver.com/someone/224349824125. Only blog.naver.com links work, and private or deleted posts are refused. Five per minute.

GET/v1/naver/forms/snapshot

When a reference post was last read, and how many blocks and photos it had. Reads our saved copy only.

NameTypeDescription
urlstringRequired. The same post link you passed to extract. Answers from our database only; Naver is not called. 404 reference_not_read when it was never read.
publishing ──► published     the channel accepted it, permalink is set
           └─► partial       some channels made it, some did not
           └─► failed        none of them made it, error says why

failed     ──► publishing   retry, and only the targets that failed go again
published  ──► cancelled    you deleted it; our record and its metrics stay

Common questions

How do I publish to several channels at once?

Name them all in accountIds on a single call. One request creates one publish with one row per channel, so you read the outcome in one place instead of correlating separate calls.

What does a 202 actually mean?

That we accepted the post and the channels have not answered yet. Read the per-target status to know what landed; the top-level response only says the work started.

Can I delete a post after it went out?

Yes. Deleting removes it from the channel and marks our record deleted, so the history still shows that it existed and when it went.