What is this?

Firstory MCP is a connector that lets you query your own show through conversation, right inside AI tools like Claude, ChatGPT, and Cursor. Once connected, you can ask the AI things like: “Which of my episodes is growing fastest lately?” “In which episodes did I mention ‘investing’?” “Compare the plays of each episode this month” — and the AI answers by reading the real data and transcripts from your Firstory dashboard in real time.

It opens up three kinds of data you can only get on Firstory for your AI to use: listening analytics, Podwise AI transcripts / content extraction, and listener comments plus Apple Podcasts reviews.

🎉 This is a first-of-its-kind podcast MCP from Firstory, currently in invite-only Closed Beta and completely free.

Before you connect

You need an AI tool that supports MCP: our top recommendation is Claude (web or desktop app); next is ChatGPT (paid plan with Developer Mode enabled); advanced users can also use Cursor / VS Code / Claude Code.
During setup, a Firstory authorization page will pop up — log in with your invited Firstory account and click “Confirm”.
After authorization, the AI only gets read-only access (it can view, but never modify or publish your show), and it can only see your own shows.
MCP server URL: https://creator-mcp.firstory.me/mcp

How to connect

Method 1: Claude (recommended, fastest)

Open Claude and go to Settings → Connectors.
Click Add custom connector.
Enter Firstory as the name and paste https://creator-mcp.firstory.me/mcp as the URL; leave the OAuth Client ID / Secret fields below empty.
Save, go back to the list, and click Connect on the connector. You’ll be redirected to the Firstory authorization page → log in with your Firstory account → click “Confirm” to authorize.
Back in the chat, check Firstory in the tools menu and start asking questions.

Method 2: ChatGPT (requires Developer Mode)

Requires a paid ChatGPT plan (Plus / Pro / Business).
Go to Settings → Connectors → Advanced and enable Developer mode.
Add a connector: name Firstory, URL https://creator-mcp.firstory.me/mcp, and choose OAuth as the authentication method.
Follow the prompts to the Firstory authorization page, log in, and approve.

⚠️ Note: ChatGPT’s regular chat and its “Deep Research connectors” mode cannot call these tools — you must use Developer Mode.

Method 3: Cursor / VS Code / Claude Code (advanced users)

Cursor / VS Code: in your MCP config file (e.g. Cursor’s mcp.json), add a server that uses streamable HTTP with these fields:

server name: firstory
type: streamableHttp
URL: https://creator-mcp.firstory.me/mcp

Claude Code: run the following command in your terminal:

claude mcp add --transport http firstory https://creator-mcp.firstory.me/mcp

With any of the above, you’ll be guided through the Firstory OAuth login on the first call.

What tools are available?

Once connected, the AI can use the following 11 tools (all read-only). You don’t need to memorize the tool names — just ask in natural language and the AI will automatically pick the right tool to answer.

Category	Tool	What it does
Shows & episodes	show_list	List all shows under your account
Shows & episodes	show_overview	Show-level overview: total plays and top episodes
Shows & episodes	episode_list	List episodes (searchable by title, paginated)
Shows & episodes	episode_detail	Full episode info and basic stats
Analytics	episode_analytics	Episode analytics overview: plays, listening platforms, listener regions
Analytics	episode_analytics_by_date	Daily play trends over a specified date range
Content extraction & search (Podwise AI)	episode_extracts_get	Get an episode’s AI extraction: summary, chapters, keywords, highlights, Q&A, transcript
Content extraction & search (Podwise AI)	episode_extracts_search	Search topics / keywords across episodes, drilling into transcripts to find exact mentions
Audience engagement	comment_list	On-platform comments on your show and episodes
Audience engagement	apple_review_list	Apple Podcasts reviews (rating, region, title)
Audience engagement	apple_review_nations	List of countries with Apple reviews

Try it now: sample questions

Once connected, drop any of these straight into the AI:

“List all my shows”
“List my 10 most recent public episodes”
“What are my show’s total plays? What are the top five episodes?”
“Put this month’s plays for each episode into a table”
“In which episodes did I mention ‘AI’? Mark the timestamps where it comes up”
“Summarize the key points and keywords of my latest episode”
“Check my recent listener comments and Apple reviews — anything I should reply to?”

Below is the actual result of typing “List my 10 most recent public episodes” in the Claude desktop app:

Level up: find your most popular topics and keywords

MCP’s biggest strength is cross-analyzing “content” against “data” — something that used to require digging through the dashboard and building spreadsheets by hand. Try this prompt:

“Read the content extractions (summaries and keywords) of my last 20 episodes, match them against each episode’s plays, analyze which ‘topics’ and ‘keywords’ are most popular, put them into a ranking, and suggest what topics I should cover next.”

The AI will automatically: ① pull each episode’s keywords and summary with the extraction tools → ② pull each episode’s plays with the analytics tools → ③ cross-reference them → ④ hand you a “high plays ↔ topics / keywords” insight report. You’ll know what your audience actually loves, and put your production effort into the right topics.

More ways to talk to your data:

“Find every episode where I mentioned a certain guest, and how those episodes performed”
“Compare my growth between my weekly and biweekly publishing periods”
“Which topics get the most listener comment engagement?”

Case study: catching “abnormal traffic” with AI

A real example: a creator noticed downloads suddenly surging on certain days in their dashboard — unreasonably high. They handed the question to the AI, which broke things down day by day and episode by episode and found that the surge was spread across 60+ old episodes that normally get almost no plays, was almost entirely unique downloads, and only happened on Apple / other RSS. Putting it together: bot traffic from a crawler sweeping the RSS feed, not real listeners. What used to take a long time of manual episode-by-episode comparison, the AI sorted out in just a few steps.

Read the full investigation here 👉 the full case study of troubleshooting unusual show analytics with AI

A few reminders

Transcripts / extractions must be generated in Studio first: if the AI says an episode “hasn’t been extracted yet”, first trigger content extraction for that episode in Firstory Studio — once it’s done, the AI will be able to read it.
Cross-episode full-text search (deep mode) is slower: to find a specific term in transcripts, the AI reads episode by episode, which can take a few to a dozen seconds — that’s normal.
Read-only throughout: all tools can currently only query; they never modify or publish your show, so feel free to explore.
Only your own shows are visible: authorization is limited to the show data under your account.

Free during Beta — and we want to hear from you

This feature will eventually become a paid premium feature on Firstory, but during the Closed Beta it’s completely free with no usage limits for invited creators. Play with it as much as you like.

After trying it, we’d especially love your feedback on three things:

What did you use it for? (which prompts / use cases)
What did it save you? (time / steps)
What’s still missing? (what else do you wish it could do)

The most-requested features go to the top of our list for the next version — your feedback directly shapes what this product becomes. Just tell us via the Firstory support chat. Thank you for being part of this brand-new experiment! 🙌

Analyze Your Show by Talking to AI: Firstory MCP (Beta Early Access)