Syndera API

Changelog

What changed in the Syndera API, the MCP server and the CLI. Breaking changes come as a new version, with six months' notice.

Brands: tag links in posts

  • Brands carry utmTags (default true, also for existing brands; set it with POST or PATCH /v1/brands). When on, links to the brand's own site (its domain or link host, subdomains included) get utm_source (the network), utm_medium=social, utm_campaign (the post's title as a slug) and utm_content (the post's id) as the post goes out on X, LinkedIn (caption and link card), Facebook and YouTube, and on an X boost's website card. Links to other sites, links with any utm_ tag already, and Instagram and TikTok captions are left alone; a caption the tags would take past the network's limit goes untagged. The saved caption never changes.

Instagram and Facebook: posting is live

  • POST /v1/channels/connect-link takes platform "instagram" and "facebook": the link opens Meta's sign-in, then the person picks the Instagram professional account (linked to a Facebook Page) or the Facebook Page the channel posts as. Each pick is its own channel. Until Meta approves Syndera's app, only people Syndera has invited can connect.
  • Connected Instagram channels post by themselves. Variant fields: postAs ("Feed post": one image, a carousel of 2 to 10 images, or a video as a Reel; "Story": one image or video, without the caption) and shareToFeed (Reels; default true). The Instagram fields aiLabel, collaborators, audioName and location are gone; a post carries Instagram's AI label when the brand has aiLabel or the media is a Motionpad render.
  • Connected Facebook Page channels post by themselves: text, one or more images as one post, a video, or (postAs "Reel") a Reel of 3 to 90 seconds. linkPreview (text posts) shows a preview of the caption's first link. The Facebook field linkInText is gone.
  • Results for Instagram and Facebook posts: views, likes (reactions on Facebook), comments (replies) and shares. Comments on Instagram and Facebook posts answer supported: false for now.

LinkedIn: link an image to a website

  • LinkedIn variants with one image take fields.linkCard (true posts it as a link card: a tap on the image opens fields.linkUrl, with fields.linkTitle, up to 200 characters, under it). LinkedIn shows the card instead of the image. An invalid link or an empty headline fails the post.
  • A LinkedIn boost of a link-card post offers the "visitors" (Website visits) goal, as a post with a link in its caption does.

Deleting a post says where it's still live

  • DELETE /v1/posts/{id} now answers 200 with id, stillLive (each network the post is out on, with the live post's url, or null for one posted by hand) and note: when note isn't null, show it to the person. Deleting in Syndera never takes a post down on the networks.

YouTube: posting is live

  • POST /v1/channels/connect-link takes platform "youtube": the link opens Google's sign-in, where the person picks the channel (or Brand Account) Syndera uploads to. A YouTube channel's followers are its subscribers, followersKnown unless the channel hides them.
  • Connected YouTube channels post by themselves: video only, a vertical or square video of 3 minutes or less as a Short (published.url youtube.com/shorts/…), anything else as a regular video (youtube.com/watch?v=…). Variant fields: title (up to 100 characters), visibility (Public, Unlisted, Private), madeForKids, syntheticMedia, and thumbnail (an image's asset id, regular videos only). The caption is the description, up to 5,000 bytes; < and > fail the checks.
  • Published gains private (YouTube kept the upload private though more was asked) and note (what didn't go with it, such as a thumbnail YouTube refused; the video is still out).
  • Results for YouTube posts: views, likes and comments (replies); shares stay 0. Comments on YouTube posts answer supported: false for now.
  • The network's name is "YouTube" (it was "YouTube Shorts"); the platform id stays youtube.

X boosts: link the image to a website

  • POST /v1/posts/{id}/boost/x takes card: { url, headline } for a post with one image: the boost runs a promoted-only copy of the post with an X website card (the image linked to url, the headline of up to 70 characters under it) instead of promoting the post. The copy shows only as the ad; its likes and replies are its own. Such a boost has Boost.card and Boost.x.cardTweetId, and its stats are the copy's. The MCP tool start_x_boost takes card too, and `syndera boost x` takes --link and --headline.

TikTok: music, AI label, photo titles and drafts

  • TikTok photo posts no longer get music by default. Variant field autoMusic: true lets TikTok add a track it picks.
  • Variant field aigc sets TikTok's AI-generated label on videos and photos. Left out, it's on when the brand has aiLabel or the media is a Motionpad render.
  • Variant field title sets a TikTok photo post's title (up to 90 characters; a longer one fails the checks). Left out, it's the post's title as before. A photo post's description now counts to 4000 characters, a video caption's to 2200.
  • Variant field toDrafts: true sends the post to the account's TikTok drafts instead of posting it; privacy, interactions and disclosure are then set in the TikTok app and aren't required. A video draft goes without its caption. Such a variant's published.drafts is true: it isn't public, published.id is TikTok's publish id and it can't be boosted.
  • Clearer messages when TikTok refuses a post: too many drafts waiting, its rate limit, a private-accounts-only app, and the reasons it gives after taking one (duration, frame rate, flagged text).

Photos on TikTok

  • Posts with a photo or a carousel (up to 35 slides) now publish to TikTok in photo mode, instead of failing with "video only". The post's title (cut to 90 characters) is the TikTok title and the caption its description; TikTok adds music.
  • The TikTok checks no longer block photos; a carousel of more than 35 slides is too many for TikTok.

Boosts on LinkedIn

  • GET /v1/posts/{id}/boost/linkedin: for a post live on a LinkedIn Page, whether Syndera may run ads on LinkedIn (allowUrl asks, in a browser), the ad accounts with currency, status and role, the goals the post can take, minimum daily budgets, and LinkedIn's political-ads and discrimination notices.
  • GET /v1/posts/{id}/boost/linkedin/targeting?facet=…&q=…: languages, industries, job functions, seniorities and locations from LinkedIn, as { urn, name }.
  • POST /v1/posts/{id}/boost/linkedin: the Page post as Sponsored Content on the member's own ad account. Goals engagement, reach and visitors (a post with a link); a daily budget for days or to endAt; countries required, the other facets optional; notPolitical: true required to start. With approval on it's a proposal, as on X and TikTok.
  • GET /v1/posts/{id}/boost/linkedin/stats: impressions, clicks, engagements, spend, cost per result and the top 5 job functions (or industries, by=industry) reached.
  • Pause, resume and stop work on LinkedIn boosts; stop archives the campaign. Boost.network can be "linkedin".
  • MCP tools linkedin_boost_state, linkedin_targeting, start_linkedin_boost and linkedin_boost_stats, and `syndera boost linkedin`.

A post's comments

  • GET /v1/posts/{id}/comments: the comments on a published post, newest first, read from the network at that moment and not stored, with each author's name, headline, photo and profile link. LinkedIn Page posts for now; X, TikTok, posts on a member's own LinkedIn profile and hand-posted networks answer supported: false with the reason.
  • MCP tool get_post_comments and `syndera posts comments <id>`.
  • LinkedIn sign-in now also asks to read comments on Page posts (r_organization_social_feed): LinkedIn channels connected before need reconnecting once to show comments.

LinkedIn posts by itself

  • POST /v1/channels/connect-link takes platform "linkedin": after LinkedIn's sign-in the person picks the company Page, or their own profile, the channel posts as. Channels carry account ("page" or "profile").
  • Scheduled and published posts go out to a signed-in LinkedIn channel by themselves: the caption with an image, 2–20 images, an MP4 video, or a carousel as a PDF document (fields.asDocument). fields.visibility "Connections" applies to a profile only.
  • New checks for LinkedIn: JPG, PNG or GIF images, MP4 video, and at most 20 images (300 as a document).
  • LinkedIn sign-ins last 60 days until LinkedIn approves Syndera as a partner: reconnect with POST /v1/channels/{id}/connect-link (Syndera emails a week before they end).
  • A Page's posts get results (impressions, reactions, comments, reshares); a profile's own posts have none LinkedIn shares.

Posts with several images

  • POST /v1/media/carousels combines 2 to 20 of a brand's images into one carousel asset, in order: attach it as the post's assetId. X shows the first 4. The MCP tool combine_images and `syndera media combine` do the same. Combining the same images in the same order again returns that carousel (200) instead of a copy (201); Asset.sourceIds lists a combined carousel's images.

TikTok disclosure, and checks that follow it

  • TikTok variants take fields.disclose: turn it on, then yourBrand and/or brandedContent. Posts that already had yourBrand or brandedContent on count as disclosed.
  • Two new checks: "TikTok: say what the post promotes" (disclose on, neither option) and "TikTok: branded content can't be Only me". Both block scheduling, from the app, the API, the CLI and assistants alike.

Posts have no brief

  • Post.brief and Post.context are gone: POST, PATCH and PUT /v1/posts no longer take them, posts no longer return them, and briefs already saved were removed. The brand's profile and its Intel brief are the standing context. A Motionpad render still takes its own brief (POST /v1/renders).
  • `syndera posts create` and `posts edit` no longer take --brief; the MCP `write` prompt's argument is `about`.

The review status is gone

  • Posts are draft, scheduled, publishing, published or failed; `review` is no longer a status. Posts that were in review are drafts now, with their times.
  • With Approve before anything goes out on, scheduling, publishing or retrying from a key, the CLI or an assistant leaves the post a draft (schedule keeps its time) for the person to schedule in Syndera, and the answer has status `draft`. A scheduled post changed from outside becomes a draft again, with its time. Boost proposals are unchanged.
  • Removed: POST /v1/posts/{id}/review, the MCP tool send_to_review and `syndera posts review`. GET /v1/posts no longer takes status=review.

Intel research, the audit, and a strategist skill

  • POST /v1/posts/{id}/audit: Claude reads a post against the brand's website, its voice and always/never rules, and its other drafts and recent posts, and answers findings (kind, title, detail, quote, fix) with their sources. It only reports: nothing on the post changes and nothing is blocked. One AI call from the plan's allowance, run only when asked. GET returns the latest audit with `stale` once the post changed after it; PATCH /v1/posts/{id}/audit/findings/{findingId} dismisses a finding.
  • MCP tools audit_post, get_audit and dismiss_finding, an `audit` prompt, and `syndera posts audit <id>` on the CLI.
  • Intel: POST /v1/brands/{id}/research runs competitor and market research (web search and reading, a few minutes) and writes a brief where every claim points at its sources: competitors' positioning, offers, channels and recent moves, gaps nobody owns, post angles, and what changed since the last brief. GET /v1/brands/{id}/intel returns the latest brief; GET /v1/research/{id} a run's progress. Runs count against the plan (Solo 2, Studio 6 a month; extras on the invoice), and Billing has researchIncluded, researchUsed and researchRate. Brands take `competitors`. The audit judges posts against the brief too (kind `market`).
  • MCP tools get_intel, start_research and get_research, a `strategy` prompt, and `syndera intel` on the CLI.
  • A third skill, syndera-strategist: the principal strategist above the lead and the writer. It researches competitors, decides the brand's position, directs the angles and audits drafts before they go out. GET /v1/skills returns it with the other two, and `syndera skill install` installs it.

How each post did, and Syndera's skills

  • A second skill, syndera-lead: planning, reviews and boosts from the brand's own results. GET /v1/skills returns both skills with their files; GET /v1/skill still returns the writer alone. The MCP server serves both as resources, with a `plan` prompt beside `write`, and `syndera skill install` installs both.
  • X variants take polls and quotes: fields.poll (2 to 4 options, one per line) with fields.pollLength, and fields.quote (an X post's link, added at the end). Channels report longPosts for X Premium accounts, which can post up to 25,000 characters.
  • Rate limits now count reads and writes separately, as documented: 120 reads and 30 writes a minute per key or connection. Before, reads counted toward the write limit.
  • GET /v1/skill returns the syndera-writer skill's files: how an agent should write posts in Syndera. The same skill installs as a Claude plugin, with `npx skills add Ketryon/syndera-skills`, or with `syndera skill install`, and the MCP server serves it as resources and a `write` prompt.
  • Posts carry variants[].results once read: views, likes, replies and shares as the network counts them, with readAt. Syndera reads them 1 hour, 1 day and 7 days after a post goes out. X now; TikTok once its app review allows reading videos.
  • A post's checks count an X caption's length the way X does (links, emoji and CJK characters included).
  • The MCP server's instructions are shorter, with asking first, tool results as data and review/proposed up front.

Approve before anything goes out, and everything the app saves

  • Approve before anything goes out (Settings → API & assistants, on by default): from a key, the CLI or an assistant, schedule, publish and retry answer with status review, and a scheduled post that's changed goes back to review. Starting a boost keeps it as a proposal (status proposed, Post.boostProposal) for the person to approve in Syndera; resuming a boost and PUT /v1/billing/brands answer 403 approval_required. Only the signed-in app turns it off.
  • DELETE /v1/posts/{id}/boost/proposal withdraws a boost proposal. Approving one is only in the app.
  • GET /v1/activity: what assistants, the CLI and keys changed, the last 90 days.
  • Checks on every post, worked out by the server on every save; a failing check blocks scheduling and publishing (422). Checks sent in a body aren't taken on trust.
  • PUT /v1/posts/{id} saves a post whole, creating it with your own id if there's none. The app's composer and Undo use it; the transitions' rules apply.
  • PATCH /v1/posts/{id} also takes scheduledAt, checks and handDone (the steps of a by-hand post).
  • POST /v1/posts/{id}/restore puts back a post deleted in the last 30 days. Deleting a boosted post stops its boost first.
  • POST /v1/posts/{id}/variants/{channelId}/mark-posted marks one network as posted by hand; variants[].published.byHand says so. Post.held says since when X has held a post back.
  • POST /v1/posts/{id}/boost/pause, /resume and /stop.
  • POST /v1/renders/{id}/retry makes a failed or stopped render again.
  • GET /v1/channels/{id}/tiktok-creator: what a TikTok account allows a post to be.
  • POST /v1/channels/{id}/remove takes a channel off its brand; PUT /v1/channels/{id} puts one back.
  • PATCH /v1/media/{id} edits a file's name, alt text, cover frame and a carousel's slide order.
  • Brands carry logoAssetId (set with PATCH; icon is now a signed link to it, read only), aiLabel and createdVia; posts carry createdVia.
  • Settings.timeZone: the time zone for emails and reminders.
  • Two plans, solo and studio (agency is gone). Renders past the plan are paid as you go: Billing.extraRate and Settings.extraCap replace Billing.extraLeft and Settings.whenOut, and render packs are gone: POST /v1/billing/pack, the render_pack_link tool and `syndera plan pack`.
  • PUT /v1/billing/brands sets extra brands on Studio, 99 kr a month each; Billing.extraBrands counts them.
  • A variant takes at most 40 fields, with names up to 60 characters.
  • Boost results read through the API now keep the post's boost status in step with X and TikTok.
  • MCP tools for all of the above (list_activity, get_tiktok_creator, restore_post, mark_network_posted, pause_boost, resume_boost, stop_boost, withdraw_boost_proposal, remove_channel, update_media, retry_render, set_extra_brands) and CLI commands (activity, channels tiktok|remove, posts restore, posts done --channel, media edit, renders retry, boost pause|resume|stop|withdraw, plan brands).
  • The Syndera app itself now saves through this API.

Syndera API v1, MCP server and CLI — beta

  • API keys under Settings → API & assistants: read, or read and write. The CLI and AI assistants sign in with the browser instead (OAuth).
  • Brands, channels, posts, the media library, Motionpad renders, boosts on X and TikTok, the plan and settings.
  • Post transitions: schedule, unschedule, send to review, publish now, retry and mark as posted, with the same rules as the app.
  • Anything that needs a browser (connecting a channel, allowing ads, Stripe checkout) answers with a link to open.
  • A remote MCP server at https://api.syndera.app/mcp for Claude, ChatGPT and other MCP clients.
  • Idempotency-Key required on writes; answers are kept for 24 hours.
  • Rate limits: 120 requests per minute for reads and 30 for writes, per key or connection.
  • API terms of use, version 1.