How to Vibe Code a Podcast Generator

You do not need to read an API reference end to end to build on Jellypod anymore. With an AI coding assistant like Claude Code or Cursor, you describe the podcast generator you want in plain language and the assistant writes the integration against the Jellypod API. This is the "vibe coding" loop: you stay in intent, the assistant handles the endpoints.

The trick that makes it reliable is the Jellypod API skill. It teaches your assistant the real endpoints, auth, and request shapes, so it builds against the actual API instead of guessing.

A concrete example

Say you keep a Notion database of article ideas. You want every new row to become a published episode without you opening the Studio. Instead of wiring that up by hand, you tell your assistant: "Watch this Notion database, and for each new row, generate a Jellypod episode from the linked URL using my two existing hosts." The assistant reads the skill, finds the right endpoints, and writes the script. You review and run it.

The workflow

1. Install the Jellypod API skill

   Give your coding assistant the API knowledge it needs:

   npx skills add Jellypod-Inc/skills --skill jellypod-api
   

   Now Claude Code, Cursor, or any skill-aware assistant knows the real endpoints and request shapes.

2. Get an API key

   Create an API key in your Jellypod account settings. The API uses the same credits as the Studio, so generating an episode through code costs the same as generating one in the app.

3. Describe the generator you want

   Tell your assistant what to build in plain language: the trigger (a new file, a webhook, a row in a sheet), the source (a URL, a PDF, pasted text), and the hosts. Be specific about the output you expect.

4. Let the assistant wire up the calls

   With the skill loaded, it pulls your hosts with GET /hosts, creates the episode with POST /episodes/generate (or POST /episodes/import if you already have a script), and polls GET /episodes/{episode_id} until the render finishes and download URLs are returned.

5. Run it and iterate

   Run the script, listen to the episode, and refine by talking to the assistant: change the hosts, shorten the episodes, add a publish step. Each pass is a conversation, not a rewrite.

What you can build

• A command that turns any URL or PDF you drop in a folder into an episode.
• A scheduled job that summarizes your team's weekly updates into a private internal show.
• A Slack shortcut that generates an episode from a thread and posts the link back.
• A content pipeline that imports finished scripts with POST /episodes/import and lets Jellypod resolve speakers and render the audio.

The point of vibe coding here is not to skip understanding your system. It is to skip the boilerplate. You describe the behavior you want, the assistant handles the API surface, and you spend your time on the part that is actually yours: what the podcast says and who it is for.

For the full endpoint list, see the API Reference.