Create a post — draft, scheduled, or published now
One call for one or many channels (`channels` takes ids or provider names, mixed). Copy goes in `content` as markdown — one field, one form. Text with no markup posts exactly as typed (handles, URLs and `snake_case` are safe); `\*` shows a literal delimiter. `scheduled_at` decides the mode: omitted saves a draft; a future instant (ISO-8601 with offset, or relative shorthand `"30m"`/`"2h"`/`"3d"`/`"1w"`) books it; `"now"` publishes immediately. Arming refuses atomically when the post is not `ready` — nothing is saved, and the 400 carries the refused post as `error.post` (`reason: "not_ready"`), so read its per-channel `errors[]` and `fixes[]` there rather than re-sending the write without `scheduled_at`. The response carries per-channel validation: each entry's `errors[]` name a `code`, the `field` to change, a human message with the real numbers, and a machine `fix` where one exists. `ready` is the server's only publishability signal — a draft with errors saves fine; only arming refuses. `autofix` (default true) applies suggested fixes server-side and reports them in `fixes_applied`.
/postsOne call for one or many channels (channels takes ids or provider names, mixed). Copy goes in content as markdown — one field, one form. Text with no markup posts exactly as typed (handles, URLs and snake_case are safe); \* shows a literal delimiter. scheduled_at decides the mode: omitted saves a draft; a future instant (ISO-8601 with offset, or relative shorthand "30m"/"2h"/"3d"/"1w") books it; "now" publishes immediately. Arming refuses atomically when the post is not ready — nothing is saved, and the 400 carries the refused post as error.post (reason: "not_ready"), so read its per-channel errors[] and fixes[] there rather than re-sending the write without scheduled_at. The response carries per-channel validation: each entry's errors[] name a code, the field to change, a human message with the real numbers, and a machine fix where one exists. ready is the server's only publishability signal — a draft with errors saves fine; only arming refuses. autofix (default true) applies suggested fixes server-side and reports them in fixes_applied.
Authorization
bearerAuth In: header
Query Parameters
Organization slug (OAuth actors with several orgs); API keys imply theirs.
Request Body
application/json
TypeScript Definitions
Use the request body type in TypeScript.
Response Body
application/json
application/json
application/json
curl -X POST "https://example.com/posts" \ -H "Content-Type: application/json" \ -d '{ "content": "string", "channels": [ "string" ], "per_channel": {}, "media": [ "string" ], "autofix": true, "labels": [ "string" ], "campaign": "string", "idempotency_key": "string", "scheduled_at": "2026-07-21T09:00:00-07:00" }'{ "id": "string", "status": "draft", "scheduled_at": "string", "ready": true, "content": { "markdown": "string", "text": "string" }, "channels": [ { "channel_id": "string", "provider": "string", "name": "string", "format": "post", "ready": true, "source": "adapted", "content": { "markdown": "string", "text": "string", "length": 0, "limit": 0 }, "errors": [ { "code": "string", "field": "content", "message": "string", "fixes": [ { "kind": "trim_to_fold", "label": "string", "max_length": 0 } ] } ], "warnings": [ { "code": "string", "message": "string", "detail": { "property1": null, "property2": null } } ], "options": { "property1": null, "property2": null }, "ink_estimate": 0, "status": "draft", "external_post_id": "string", "url": "string", "error": "string", "published_at": "string" } ], "media": [ { "id": "string", "ready": true, "failed": true, "alt_text": "string", "source_url": "string" } ], "labels": [ { "id": "string", "name": "string" } ], "campaign": "string", "approval": { "required": true, "state": "not_requested" }, "fixes_applied": [ { "channel_id": "string", "provider": "string", "kind": "string" } ], "ink_estimate": 0, "created_at": "string", "updated_at": "string", "published_at": "string"}{ "error": { "code": "string", "message": "string", "reason": "string", "index": 0, "request_id": "string" }}{ "error": { "code": "string", "message": "string", "reason": "string", "index": 0, "request_id": "string" }}List posts (keyset paginated) GET
Newest first, keyset-paginated via `next_cursor`. Narrow with `statuses`, a scheduled-time range, `campaign`, `label`, or `channel` (a channel id or a provider name like `linkedin`) — so "my scheduled LinkedIn posts" is one call, not a range pulled down and filtered locally. Rows are excerpts and carry no fit verdict unless you ask: `include=ready` adds `ready` per row — absent means "not checked", which is not the same as `false` — and caps the page at 25, because each judged row then costs a full post read. Read one post in full with `GET /posts/{id}`.
The post, its per-channel validation, and live publish results GET
One read: per-channel content with counted length and provider limit, `errors[]` and `warnings[]`, media readiness, approval state, and — after fan-out — each channel's publish result (`url`, `external_post_id`, or a sanitized `error`). Poll this after publish; there is no separate status call. The response's `ETag` is the post's `updated_at`: send it back as `If-Match` on a write to lose safely (409) if someone edited in between.