Start With a Meta Muse Connector for LinkedIn Posting

Building a Meta Muse connector for LinkedIn posting means filling in no platform configuration at all, because the AdaptlyPost spec carries no LinkedIn config object. No postType, no privacy enum, no page id. Text, media and a connection id, and the post ships. That absence is exactly why LinkedIn is where you point an agent whose code you have not read.

The short version: build the connector once from the public spec the way the Meta Muse connector guide sets out, then send LinkedIn one text-only post and read the per-platform result back. If that round trip works, the connector works.

Why is LinkedIn the network to test a Muse connector on?

LinkedIn is the network to test on because the AdaptlyPost spec defines no LinkedIn config object, so a post that fails points at the connector rather than at a platform field you filled in wrong. Muse writes the client code itself and Meta does not review it, so the first LinkedIn post is the only check that code gets. Send it somewhere with nothing else to blame.

What does a LinkedIn post need in the AdaptlyPost API?

A LinkedIn post needs three required values plus the copy: platforms set to [LINKEDIN], contentType, timezone, and the text itself. linkedinConnectionIds from GET /social-accounts picks which LinkedIn account it lands on. With contentType: TEXT there is no media, so the connector skips POST /upload-urls entirely and the whole publish is a single call.

POST /social-posts
  platforms:    [LINKEDIN]
  contentType:  TEXT
  text:         "We shipped per-platform captions this week. One call,
                 nine networks, different copy on each."
  scheduledAt:  "2026-09-22T08:00:00"
  timezone:     "Europe/Kyiv"

Compare that against what the same connector has to get right on TikTok or Pinterest and the reason for starting here is obvious. Every extra enum is another value Muse fills in from a spec it read once, and a wrong enum fails at publish time rather than at request time. LinkedIn gives it nothing to guess.

What do I paste into Muse for a LinkedIn connector?

You paste the public spec URL and an instruction to read before it writes. Parallel's hands-on test of 14 September 2026 found no "add MCP server" setting in Muse, so as of 20 September 2026 the route in is a Custom Connector built from an OpenAPI document, and the full build, credential flow and revocation path live in the pillar guide. The LinkedIn-specific part is only the last line.

Build a custom connector for AdaptlyPost. The OpenAPI spec is at
https://post.adaptlypost.com/post/api/v1/openapi.json. It is public, so read
it without logging in.
 
Auth is a bearer token I will paste into the secure credential prompt.
 
Do not publish anything yet. Call GET /social-accounts and show me my
LinkedIn connections and the publishing operations you found.

What does Muse still get wrong on LinkedIn?

Muse gets the tone and the cadence wrong on LinkedIn, not the fields. There is nothing to misconfigure, so every remaining failure is editorial, and LinkedIn punishes editorial failure harder than the feeds do because the audience is people you work with.

Three habits show up in the drafts Muse hands back before you put them in platformTexts. It opens with an announcement verb and a superlative. It adds three hashtags, which read as automation to anyone scrolling. And it posts on a schedule so even that the account starts to look like a bot, because a routine with no exit clause invents a post on a quiet week.

Use platformTexts to give LinkedIn its own copy in the same call. A 1,200-character LinkedIn version and a 200-character X version go out together, and omitting the field sends identical text everywhere, which sends one length to all nine networks and reads badly on eight of them. The same split is what makes the Grok bot template for LinkedIn posting worth reading next, since it runs the cadence side of this from a saved routine.

Can Muse keep a weekly LinkedIn cadence going?

Muse keeps working after the app is closed and comes back when it needs approval, but as of 20 September 2026 there is no public API for driving your Muse from outside, so a cadence cannot be handed to a teammate or checked into a repo. Meta publishes no answer for whether a connector can be shared with a teammate, and its in-product warning covers every copy: "Muse may take unexpected actions. Monitor it carefully."

The durable version of a cadence lives on the AdaptlyPost side instead. POST /social-posts/bulk takes 1 to 100 posts in one call, each with its own scheduledAt under one timezone for the whole batch, so a month of LinkedIn updates becomes a single approval rather than four weekly ones. Ask Muse to draft the batch, read it yourself, then let it send the bulk call once. The queue then runs whether or not Muse is running.

What breaks on LinkedIn specifically?

Very little breaks at the field level, which is the honest answer, and the failures that do happen are account-level. An expired LinkedIn connection returns an auth error at publish time, not at draft time, so the agent reports a scheduled post and the post never appears. A personal profile connected where a Company Page was intended publishes successfully to the wrong place, and GET /social-posts/{id}/results returns that success with the wrong account name attached, which is the only place the mistake shows.

Two Muse-specific failures sit on top of that. A custom connector is saved as a reusable skill, and as of 20 September 2026 Meta publishes no answer for whether that saved skill survives an endpoint change. And on that same date it publishes no answer for what Muse does when your API returns 429, so a bulk call that trips a rate limit has undefined behaviour on the agent's side. Both arguments point the same way: call GET /social-posts/{id}/results after every publish and read the account name in the response, rather than trusting the agent's summary of what it did.

Frequently Asked Questions

Can Muse post to LinkedIn without a custom connector?

No. Meta connects Facebook, Instagram and Threads to Muse automatically when the accounts share an Accounts Centre, and LinkedIn appears neither in Alexandr Wang's 8 September 2026 connector list nor in the launch roster Parallel published on 14 September 2026. Without a connector the only route left is the browser sub-agent driving linkedin.com, which breaks the week a button moves.

Does a LinkedIn post through Muse need an image?

No. LinkedIn accepts a text-only post with contentType: TEXT and nothing in the media fields, so the connector skips POST /upload-urls entirely and the whole publish is a single call. That is also what makes it the cleanest first test of a new connector.

Can the connector post to a Company Page and a personal profile?

Yes. Both appear as separate connections in GET /social-accounts, and the LinkedIn connection field takes an array, so one call reaches both. Ask Muse to read the account names back before it publishes, because the connection ids do not say which is which.

Where does the AdaptlyPost token go?

The token goes into Muse's secure credential prompt, never into a chat message. The pillar guide has the mechanism: the agent's runtime only ever holds a surrogate, and Sentinel swaps in the real credential at the network boundary. Create the token at adaptlypost.com/api-tokens.

Does a LinkedIn connector cost extra on either side?

No. Muse bills a LinkedIn connector out of the same weekly token meter as anything else it does, and the pillar guide has the Power and Max figures. On the AdaptlyPost side, API access is included on Creator at $19 a month, Pro at $39 and Enterprise at $89, and on the 7-day free trial.

What fields does a LinkedIn post need in a Muse connector?

Only the shared ones: platforms set to [LINKEDIN], contentType, timezone and the text, plus linkedinConnectionIds to pick the account. The AdaptlyPost spec has no LinkedIn config object, so there is no platform field for Muse to invent.

Can Muse write different copy for LinkedIn and X in one call?

Use platformTexts to give each network its own text in the same POST /social-posts call. A 1,200-character LinkedIn version and a 200-character X version go out together. Leave the field out and every network gets the same text.

Why does my LinkedIn post show as scheduled but never appear?

Most likely the LinkedIn connection lapsed after the post was scheduled: it is accepted at draft time and returns an auth error at publish time, so the agent reports a scheduled post and nothing goes live. A connection already marked unauthorized is refused with a 400 on the POST /social-posts call itself. Check the connection, then call GET /social-posts/{id}/results.

How many LinkedIn posts can I schedule in one call?

POST /social-posts/bulk takes 1 to 100 posts in one call. Each post has its own scheduledAt, and the whole batch shares one timezone. A month of LinkedIn updates becomes a single approval.

How do I check a Muse post went to the right LinkedIn account?

Call GET /social-posts/{id}/results after every publish and read the account name in the response. A personal profile connected where a Company Page was intended publishes successfully to the wrong place, and that response is the only place the mistake shows. Do not rely on the agent's summary.

Create an AdaptlyPost API token and hand Muse the spec, or read the full connector guide before you build one.

Sources: Meta's Muse connectors help article, Meta's safety engineering post, Parallel's custom integrations write-up (14 September 2026), and the AdaptlyPost OpenAPI spec. Checked 20 September 2026.