Contents

API reference / Review

The posts to review (Good / Not relevant)

GET /api/v1/review/queue

list_review_queue Read MCP tool

The kept posts a person should rule on, from the feed (mentions) and from target accounts. `slice` picks the list: needs_review (default) is the posts the AI reviewer's two readings or the ranking model disagree about, then the top of the rest; all is every unreviewed kept post, strongest first; suggested_good and suggested_not_relevant are what the AI proposed; reviewed is what people already ruled on, newest first (to undo). Each item has the quote, its source, age and strength, the AI's `suggestion` (verdict, the two readings, a one-line reason, `flag`: agree, disagree or dont_miss) and why it is in the queue. Rule on an item with update_mention (a mention) or update_account_post (an account's post), `verdict` good or not_relevant: a human verdict is final, and the reviewer and the ranking model learn from it. `counts` says how much is waiting, how many verdicts there are and which ranking model the workspace scores with (its own once it has about 200 verdicts, a pooled one before). Reading costs nothing.

Any API key or OAuth token, or a signed-in member.

Parameters

NameTypeDescription
slice
query
string needs_review (default), all, suggested_good, suggested_not_relevant or reviewed.
One of: needs_review all suggested_good suggested_not_relevant reviewed
project
query
string Only this project's mentions.
account
query
string Only this account's posts (its domain).
kind
query
string Only mentions, or only target accounts' posts.
One of: mention account_post
limit
query
integer
1–100
How many items (default 20).
offset
query
integer
at least 0
How many to skip (next_offset from the last page).

Returns

{ data: QueueItem[], total, next_offset, counts: { unreviewed, suggested, disagreements, verdicts, examples_pass, model } }

Example

curl "https://sniffington.com/api/v1/review/queue" \
  -H "Authorization: Bearer $SCOUT_API_KEY"