Model Context Protocol is an open standard for connecting AI assistants to outside tools and data. Adding our server gives your assistant five tools it can call while answering you, so it quotes real sponsorship data instead of guessing from memory.

Yes. A free Track Podcasts account includes 25 MCP tool calls a day, resetting at midnight UTC. No card, and connecting does not start a trial or use one up.

What does a paid plan add?

Higher daily limits, the full list of shows for a brand rather than the first six, and the receipts: the verbatim ad read, its promo code and its timestamp in the episode. Paid plans use the same server and the same key.

Which assistants work with it?

Anything that speaks MCP over HTTP: Claude (desktop, web and Claude Code), Cursor, and ChatGPT via a custom connector. The configuration below is the same standard JSON every one of them accepts.

Where does the data come from?

From the audio. We transcribe podcast episodes and detect the ad reads in the transcript, so a brand is listed only when the ad actually ran — not because it appears on a media kit or a rate card. How we do it is on the methodology page.

Do you cover every podcast?

No, and no one does. We index the podcast catalogue and transcribe continuously, prioritising active shows. Coverage is growing and is not complete, so a show or brand missing from an answer may simply not be indexed yet. Every response carries that caveat and a get_coverage_info tool exists so your assistant can check the real numbers before it answers.

Can the assistant change anything in my account?

No. Every tool is read-only. The server has no tool that writes, deletes or spends anything.

What happens when I run out of calls?

The tool returns a clear limit message your assistant will relay, and the count resets at midnight UTC. Nothing is charged and nothing is cut off.

Free MCP server

Ask your AI assistant who sponsors any podcast

Connect Track Podcasts to ChatGPT, Claude or Cursor and your assistant can look up real podcast sponsorship data mid-conversation — every brand detected on a show, every show a brand buys, and how that has moved month to month.

Get a free API key How to connect

25 calls a day free · No card · Connecting does not start a trial

What this is

Ask an assistant who sponsors a podcast today and it will answer from whatever it happened to read during training — usually a year stale, often a guess, never sourced. It has no way to check.

The Model Context Protocol fixes that by letting an assistant call out to a live source while it is answering. This server is ours: five read-only tools over the sponsorship data behind the Advertiser Index, detected from transcribed audio rather than from media kits.

Every answer comes back with a link to the page the data is on, a date, and a note about what our coverage does and does not include — so the assistant can cite it, and can tell you when it does not know.

The five tools

Your assistant picks whichever one fits the question. You never name them.

get_show_sponsors

Sponsors of a show

Every brand whose ad read was detected on that show, how many reads, whether they were host-read or network-inserted, and when each was first and last heard.

“Who sponsors the Glenn Beck Program?”

get_brand_podcasts

Where a brand advertises

The shows a brand has been heard on, newest first, with the exact total. Takes a name, a slug or a domain — a domain is the most reliable.

“Which podcasts does Squarespace advertise on?”

get_sponsor_activity_summary

A brand over time

Shows reached, ad reads and organic mentions per month, up to 36 months back, from a table we rebuild daily.

“Has AG1 been buying more podcast ads this year?”

get_coverage_info

What the data covers

Shows indexed, episodes transcribed, and the real number of brands and shows with detected ads — so the assistant can qualify its own answer instead of overstating it.

“How complete is this data?”

get_busiest_categories

Busiest categories

The categories with the most detected ad reads over the last 30, 90 or 365 days, with the brands and shows behind each. Aggregate only.

“Which advertiser categories are buying most right now?”

More example prompts on podcast sponsor data for AI.

How to connect

Create a free account, make a key in Settings → API, then add the server to your assistant. Two minutes.

1
Create a free account
No card. The free plan includes MCP access; signing up does not start a trial.
2
Make a key
Settings → API. Keys start with tp_live_ and are shown once — we store only a hash.
3
Add the server
Paste the config below into your assistant, with your key in the Authorization header.

Claude Desktop, Cursor and most others

mcp.json

{
  "mcpServers": {
    "trackpodcasts": {
      "type": "http",
      "url": "https://trackpodcasts.com/api/mcp",
      "headers": {
        "Authorization": "Bearer tp_live_your_key_here"
      }
    }
  }
}

Claude Desktop: Settings → Developer → Edit config. Cursor: .cursor/mcp.json in your project, or the global one in Settings → MCP.

Claude Code

Terminal

claude mcp add --transport http \
  trackpodcasts https://trackpodcasts.com/api/mcp \
  --header "Authorization: Bearer tp_live_your_key_here"

ChatGPT: Settings → Connectors → add a custom connector pointing at https://trackpodcasts.com/api/mcp and supply the same Authorization header. Availability of custom connectors depends on your ChatGPT plan.

Prefer to call it yourself? The same data is on the REST API.

Free and paid

Same server, same key. A paid plan raises the ceiling and unlocks the receipts.

	Free	Pro and above, from $30/mo
Tool calls a day	25	100 on Pro, more on Team and Agency
All five tools	Yes	Yes
Brands on a show	Every one we detect	Every one we detect
Shows for a brand	First six, with the exact total	Every show
The verbatim ad read	No	Quote, promo code and timestamp
Monthly activity and market aggregates	Yes	Yes
CSV export on the website	No	Yes

Compare every plan

What the data does and does not cover

Sponsorships are detected in transcribed audio. A brand is listed only when the ad was actually heard in an episode we transcribed.
Coverage is growing and is not complete. A show or brand missing from an answer may simply not be indexed yet — absence is not evidence a sponsorship never happened.
Network-inserted ads are served per download, so what we transcribed is one instance at one moment and may not be what another listener heard. Host-read and network-inserted reads are always reported separately, never blended.
The get_coverage_info tool returns the current figures so your assistant can qualify its own answer. The same numbers are public on the stats page and the method is written up under methodology.

Questions

What is MCP?: Model Context Protocol is an open standard for connecting AI assistants to outside tools and data. Adding our server gives your assistant five tools it can call while answering you, so it quotes real sponsorship data instead of guessing from memory.
Is it really free?: Yes. A free Track Podcasts account includes 25 MCP tool calls a day, resetting at midnight UTC. No card, and connecting does not start a trial or use one up.
What does a paid plan add?: Higher daily limits, the full list of shows for a brand rather than the first six, and the receipts: the verbatim ad read, its promo code and its timestamp in the episode. Paid plans use the same server and the same key.
Which assistants work with it?: Anything that speaks MCP over HTTP: Claude (desktop, web and Claude Code), Cursor, and ChatGPT via a custom connector. The configuration below is the same standard JSON every one of them accepts.
Where does the data come from?: From the audio. We transcribe podcast episodes and detect the ad reads in the transcript, so a brand is listed only when the ad actually ran — not because it appears on a media kit or a rate card. How we do it is on the methodology page.
Do you cover every podcast?: No, and no one does. We index the podcast catalogue and transcribe continuously, prioritising active shows. Coverage is growing and is not complete, so a show or brand missing from an answer may simply not be indexed yet. Every response carries that caveat and a get_coverage_info tool exists so your assistant can check the real numbers before it answers.
Can the assistant change anything in my account?: No. Every tool is read-only. The server has no tool that writes, deletes or spends anything.
What happens when I run out of calls?: The tool returns a clear limit message your assistant will relay, and the count resets at midnight UTC. Nothing is charged and nothing is cut off.

Give your assistant the sponsor data

A free account, a key, and one block of config. 25 calls a day, no card.

Get a free API key