MCP Server
Podwise runs a hosted MCP (Model Context Protocol) server, so you can use Podwise in Claude, ChatGPT, Cursor, and other MCP clients. Search episodes, read summaries and transcripts, ask questions, create Clips, and send notes to your apps, all from a chat.
In Claude and ChatGPT, search for Podwise in the plugin directory and add it. In other MCP clients, add the server URL:
https://mcp.podwise.ai/mcpThe server uses the Streamable HTTP transport. Clients sign in with your Podwise account through OAuth, so all you do is approve access.
Requires a Pro or Enterprise plan. The MCP server follows the same access rules and rate limits as the Open API. Free accounts can connect, but every tool call returns a notice that the feature requires a Pro or Enterprise plan, with a link to the plan comparison page. Upgrade your plan.
Connect your AI app
Podwise is listed in the Claude and ChatGPT plugin directories. The Podwise plugin includes the MCP server and the Podwise skill, which adds ready-made workflows such as catching up on new episodes, writing a weekly recap, taking episode notes, and researching a topic across podcasts.
Works on claude.ai and Claude Desktop.
Add the Podwise plugin (Pro, Max, Team, and Enterprise plans)
- Open Customize in the left sidebar, go to Plugins > Discover, and search for
Podwise. - Select Podwise and click Add.
- On the plugin page, switch to the Connectors tab and click Connect. Sign in to Podwise if needed and click Allow.
- Ask something like "Catch me up on new episodes from the podcasts I follow", or type
/to pick a Podwise workflow.
Plugins you add are saved to your Claude account, so Podwise is also available in Cowork and Claude Code.
Add the Podwise connector only (all plans)
If plugins are not available on your plan, add the connector instead. It provides the same tools without the Podwise skill.
- Open Customize > Connectors, click + next to Connectors, and search for
Podwise. - Select Podwise and click Connect. Sign in to Podwise if needed and click Allow.
- In a chat, enable Podwise from the + button under Connectors.
On Team and Enterprise plans, an Owner first adds Podwise in Organization settings > Connectors > Browse connectors. Members then find it in Customize > Connectors and click Connect.
Add Podwise as a custom connector
Use this if your organization does not allow connectors from the directory.
- Open Customize > Connectors, click +, then choose Add custom connector.
- Name it
Podwise, pastehttps://mcp.podwise.ai/mcpas the URL, and click Add. Leave the OAuth fields in Advanced settings empty. - Click Connect. Claude opens the Podwise authorization page; sign in if needed and click Allow.
Claude Free accounts are limited to one custom connector.
Use an API access token
Developer tools, scripts, and agents that run without a browser can skip OAuth and send an API access token in the Authorization header. Create a token in Settings > Developer. The same token works with the Open API.
Authorization: Bearer YOUR_API_TOKENclaude mcp add --transport http podwise https://mcp.podwise.ai/mcp \
--header "Authorization: Bearer YOUR_API_TOKEN"Treat the token like a password. Anyone with it can act on your Podwise account; disable or delete it in Settings > Developer if it leaks.
Manage connected apps
Every AI app you authorize through OAuth, including the Podwise plugin and connector, appears in Settings > Connections > Connected Apps, with the date it was connected and when it was last active. Click Disconnect to revoke its access immediately. You can connect it again from the app at any time.
Apps connected with an API access token do not appear there; manage those tokens in Settings > Developer.
Example prompts
Once the server is connected, ask in plain language. The skill picks the tools; with a plain MCP connection, the model chooses them from the tool list below.
| Prompt | What the agent calls |
|---|---|
"Summarize this episode: https://app.podwise.ai/dashboard/episodes/<seq>" | get_episode, then get_episode_summary |
| "Catch me up on this week's episodes" | list_followed_episodes, then summaries for the ones that matter |
| "What do podcasts say about developer hiring?" | search_episodes, then summaries and transcripts from the results |
| "Show me the transcript of this episode with timestamps" | get_episode_transcript, page by page |
| "Save the argument at 12:30 as a Clip" | get_episode_transcript, then create_clip |
| "Send my notes on this episode to Notion" | send_episode (Notion must be connected in Podwise first) |
| "Translate this episode into Japanese" | translate_episode, then list_episode_translations |
| "Process this YouTube video" | import_episode, then process_episode (asks for confirmation; uses credits) |
Some prompts to start with:
Summarize this episode and list its chapters with timestamps:
https://app.podwise.ai/dashboard/episodes/<seq>Find every episode in my followed podcasts from the last two weeks,
and tell me which three are worth reading today — one line each.Ask Podwise: what do podcasts say about AI agents replacing junior
developer work? Cite the episodes you used.Episode and podcast links are accepted in chat; the tools themselves take the numeric seq from the link.
Tools
Episodes and podcasts are identified by their numeric seq, the number in Podwise URLs such as https://app.podwise.ai/dashboard/episodes/<seq>.
Discovery
| Tool | Description |
|---|---|
search_episodes | Search episodes by keywords |
search_podcasts | Search podcasts by keywords |
get_popular_episodes | List episodes currently popular on Podwise |
list_podcast_episodes | List a podcast's episodes within a date range |
Episodes
| Tool | Description |
|---|---|
get_episode | Get episode details and AI processing status |
get_episode_summary | Get the AI summary: overview, takeaways, chapters, and highlights by default, plus Q&A, keywords, and mind map on request |
get_episode_transcript | Get the transcript with timestamps and speakers, page by page |
process_episode | Start AI processing for an episode (uses AI processing credits) |
translate_episode | Request an AI translation of a transcribed episode |
list_episode_translations | List an episode's translations and their status |
import_episode | Import a single episode from a Xiaoyuzhou episode, Apple Podcasts episode, or YouTube video URL (podcast and channel URLs are rejected) |
export_episode_markdown | Export an episode as a Markdown note, optionally for Obsidian or Logseq (long exports return a download link) |
export_episode_srt | Export an episode's transcript as SRT subtitles, optionally translated or bilingual (long exports return a download link) |
send_episode | Send an episode to your connected Notion database or Readwise Reader |
set_episode_read | Mark an episode as read or unread |
get_episode_summary accepts an optional sections array with any of overview, takeaways, chapters, highlights, qa, keywords, and mindmap. Without it, the tool returns the overview, takeaways, chapters, and highlights; the Q&A, keywords, and mind map are returned only when requested. The response lists the sections it left out in omittedSections.
Library and Ask
| Tool | Description |
|---|---|
get_me | Get your account email, plan, and remaining AI processing credits |
list_followed_podcasts | List followed podcasts with recent new episodes |
list_followed_episodes | List new episodes from podcasts you follow |
list_my_episodes | List episodes you recently played or marked as read |
set_podcast_follow | Follow or unfollow a podcast |
ask_podwise | Ask a question answered from podcast transcripts, with cited sources (counts against your Ask quota) |
get_ask_history | List your past questions or get one full answer |
Clips
| Tool | Description |
|---|---|
list_clips | List your recent Clips or the Clips of one episode |
get_clip | Get a Clip with its content, takeaways, and media URL |
create_clip | Save a key moment of an episode as a Clip |
delete_clip | Permanently delete a Clip |
export_clips_markdown | Export an episode's Clips as Markdown (long exports return a download link) |
send_clips | Send Clips to your connected Notion database or Readwise |
Audio uploads
| Tool | Description |
|---|---|
start_audio_upload | Start uploading a local audio file and get the upload links |
complete_audio_upload | Create the episode from the uploaded file and start AI processing (uses AI processing credits) |
Notion, Readwise, and Readwise Reader must already be connected in Podwise Settings > Connections before send_episode or send_clips can use them.
Enterprise tools
Accounts with an active Enterprise entitlement also see these tools, which wrap the Enterprise API. Usage is billed to your Enterprise organization. If your Enterprise access changes, reconnect the app to refresh its tool list.
| Tool | Description |
|---|---|
enterprise_process_audio | Submit an audio URL or an uploaded file for processing |
enterprise_get_status | Get processing or translation status for up to 100 items |
enterprise_get_result | Get the processing result of one item |
enterprise_query_results | Get selected result sections for up to 50 items |
enterprise_translate | Request a translation of a processed item |
enterprise_get_translation | Get a finished translation |
enterprise_get_usage | Get processing history and usage for a billing cycle |
enterprise_export | Export a result as Markdown, SRT, CSV, or XMind (XMind and long exports return a download link) |
Long exports
export_episode_markdown, export_clips_markdown, export_episode_srt, and enterprise_export return short exports inline. Long exports return a download link instead, valid for 10 minutes, and XMind exports always return a link. The cutoff is based on the estimated size in tokens, so Chinese, Japanese, and Korean content switches to a link sooner than English.
Confirmation prompts
process_episode, complete_audio_upload, and enterprise_process_audio are marked as destructive because they consume AI processing credits or, for the Enterprise tool, are billed to your Enterprise organization. delete_clip is marked the same way because it permanently deletes a Clip. Apps such as Claude and ChatGPT ask you to confirm before every call to these tools.
Uploading local audio
A remote MCP server cannot read files on your computer, so audio files are uploaded outside the chat:
- The AI app calls
start_audio_upload, which returns an upload URL and a browser upload link. Both expire after one hour. - Agents that can run shell commands, such as Claude Code, upload the file directly. In chat apps like Claude and ChatGPT, open the browser link and choose the file yourself.
- The AI app calls
complete_audio_uploadto create the episode and start AI processing. This uses AI processing credits, so the app asks you to confirm first.
Hosted and local MCP servers
The Podwise CLI also includes a local MCP server started with podwise mcp. Both give AI apps access to Podwise with the same plan requirements; they differ in where they run and how they sign in.
| Hosted MCP server | Local MCP server | |
|---|---|---|
| Setup | Add Podwise from the Claude or ChatGPT plugin directory, or paste https://mcp.podwise.ai/mcp | Install the Podwise CLI, then run podwise mcp |
| Sign-in | OAuth in your browser, or an API access token | podwise auth |
| Works with | Claude, ChatGPT, and any remote MCP client, including web and mobile apps | Desktop clients that can launch local commands |
| Best for | Most users | Developers already using the CLI |