- “Who sponsors TBPN right now, and since when?”
- “Which brands sponsor both TBPN and the All-In Podcast?”
- “Where has Ramp run ad reads this quarter, and with which promo codes?”
- “Which shows recommend Figma without being paid to?”
Ask your agent who sponsors a show
Connect the MCP server, then paste one of these. On any plan, the answer names a show’s top 3 sponsors by ad-read count, or up to 3 of a brand’s top shows by ad-read count, with the newest ad read your plan can open for each, its timestamp, and the totals.arcmira.sponsors(channelId) or arcmira.recommendations(entityId) through arcmira_execute_read, the two reads below. Below Pro+, access.message says how many sponsors or shows and ad reads the answer left out. Pro+ lists every one with its timestamp.
Find who sponsors a show
GET /v1/channels/{channel_id}/sponsors ranks a show’s recurring sponsors by ad reads. channel_id is the YouTube UC... id; resolve a show by name with type=channel to get it.
GET /v1/channels/{id}/sponsors (arcmira.sponsors over MCP) returns the show’s total sponsors and ad reads and its top 3 sponsors by ad-read count, each with one receipt: the newest ad read your plan can open, with its timestamp:
meta.total and meta.ad_reads include reads your plan cannot list yet, and access says how many sponsors and ad reads the slice left out. Sponsors with the same number of ad reads are ordered by name, then by id, so the same three come back on every call. Below Pro+, a show can also return its totals with no rows: sponsors is empty, and access says how many sponsors and ad reads Pro+ lists.
Pro+ lists every sponsor, each row without receipt, and the response has no access and no meta.ad_reads. Pro+ also takes the filters status (active, lapsed, ended, uncertain), min_ad_reads (default 3) and limit. Sending a filter below Pro+ answers 403 filter_requires_paid.
Find which shows a brand sponsors
GET /v1/recommendations?entity_id=ent_14 lists each moment a brand was promoted, newest first, with the quote. Pro+ gets every row.
On a plan below Pro+, GET /v1/recommendations (arcmira.recommendations over MCP) returns the brand’s total shows and ad reads and up to 3 of its top shows by ad-read count in shows, each with one receipt: the newest ad read your plan can open, with its timestamp. recommendations is empty:
meta.total and meta.ad_reads include reads your plan cannot list yet. Shows with the same number of ad reads are ordered by name, then by channel id. Below Pro+, a brand can also return its totals with fewer than three shows, or none: shows is shorter or empty, and access says how many shows and ad reads Pro+ lists. class=sponsored returns the same slice, and a limit under 3 shortens it. Paging, after, before and the other filters stay Pro+: a filter answers 403 filter_requires_paid and a cursor answers 403 pagination_gated.
recommendations looks like this:
Filter with
class (omit it for all three), channel_id, min_confidence (default 0.7) and the after/before window from Search. Rows whose class is disputed are left out unless you pass include_disputed=true.
Every row is checked against the transcript. An ad read whose words and brand never appear in the video is not served, and sponsor counts leave it out.
Cost
On Pro+, each sponsor and each commercial row costs 40 credits; the free slice costs nothing. See Plans and usage.Wrong classification
A false ad read hurts a brand, and a missed one hurts a negotiation. Report it with typerecommendations and the row’s com_ id, or channel_sponsors with the sponsor’s ent_ id and the channel. Reviewing commercial rows needs Pro+, including rows from the free slice.