Quickstart
You need a Podwise account before anything else. Sign in at app.podwise.ai with Google, Apple, GitHub, email and password, or a magic link. Use the same method on every device; a different method with a different address creates a second account.
Then pick the path that matches what you want to do. Each one takes about five minutes.
Web app
This is the path most people want: read a podcast instead of listening to it.
- Open app.podwise.ai and sign in.
- Search for a show you listen to and open one of its episodes. Podwise indexes podcasts from Apple Podcasts, so most public shows are already there.
- Read the episode. The Summary tab has the overview, takeaways, and chapters; Transcript has the full text with speaker labels and timestamps, and clicking a line jumps to that moment in the audio. The Mindmap tab shows the episode's structure.
- Every plan starts with a 7-day trial, so you can process something new: open an episode that isn't processed yet, click AI Process, and confirm. One podcast episode or YouTube video uses 1 AI processing credit.
- Save a moment as a Clip from the player, or click Export on the episode to send it to Notion, Readwise Reader, Obsidian, or Logseq. Those targets need to be connected first in Settings › Connections.
Opening an episode that is already processed — including most popular episodes — never uses credits; only processing a new one does.
Done when you have a summary open and know where the transcript, mind map, and clips are.
Browse and search work without a subscription. Processing new episodes and opening AI content you haven't unlocked need a plan, and every plan starts with a 7-day trial. See Plans, Credits & Billing for the details.
AI app
Use this if you want to ask Claude, ChatGPT, Cursor, or Codex about podcasts.
-
Add Podwise to your AI app. In Claude and ChatGPT, install the Podwise plugin from the plugin directory. In Cursor, VS Code, Codex, or another MCP client, add the server URL:
https://mcp.podwise.ai/mcp -
Complete the sign-in when the app asks. OAuth runs in your browser, so there is no key to copy.
-
Ask for something:
Summarize this episode: https://app.podwise.ai/dashboard/episodes/<seq> Catch me up on this week's episodes from the podcasts I follow.
The full per-client steps are in MCP Server, and the plugin's skill adds named workflows such as catch-up and weekly recap — see Skills & Plugins.
Done when the agent returns a summary with timestamps, without you pasting a token.
The MCP server requires a Pro or Enterprise plan. A free account can connect, but tool calls return a notice that the feature needs a paid plan.
CLI
Use this if you want Podwise in a script, a cron job, or an agent that runs commands.
-
Install the
podwisecommand (Homebrew on macOS, an install script, a binary, or from source):brew install hardhackerlabs/podwise-tap/podwise -
Sign in. This opens your browser and stores the key for you:
podwise auth podwise config show -
Fetch something:
podwise search episode "machine learning" --limit 5 podwise get summary https://app.podwise.ai/dashboard/episodes/<seq> -
Optional: run it as a local MCP server with
podwise mcp, so desktop AI apps can launch it themselves.
Done when a summary prints in your terminal. The command list, output notes, and the local MCP config are in CLI.
The CLI requires a Pro or Enterprise plan. Processing an episode with podwise process spends AI processing credits like any other processing request.
What just happened
Whatever path you took, the result is the same object: one episode with its AI content attached — summary, transcript, mind map, highlights, keywords, and any Clips you saved. The web app shows it as tabs; the MCP server and the API return the same sections as separate fields.
That is the vocabulary the rest of these docs use. Core Concepts defines each term in one place.
Next steps
- Core Concepts: episode, processing, summary sections, credits, Clips.
- Authentication: which credential to use where.
- Recipes: ready-made jobs, from a thread to a weekly recap.
- FAQ: plans, limits, imports, and account questions.