CLI

The podwise command is the terminal version of Podwise, built for scripts and for agents that can run commands. It searches podcasts, fetches an episode's AI content, processes new episodes, and asks questions — without a browser.

It is also the local option in the AI tool set: podwise mcp runs Podwise as a local MCP server over stdin/stdout, for desktop AI apps that launch commands themselves.

InstallCopy Anchor Link

brew install hardhackerlabs/podwise-tap/podwise

Check the installation with podwise --help; it lists every command and flag for your version.

Sign inCopy Anchor Link

podwise auth          # opens the browser and saves an API access token
podwise config show   # prints the account the CLI is using

The configuration lives at ~/.config/podwise/config.toml. If you already have a token from Settings › Developer, set it directly:

podwise config set api_key your-sk-xxxx

Common commandsCopy Anchor Link

Process anythingCopy Anchor Link

podwise process accepts the same sources as the web app: a Podwise episode URL, a Xiaoyuzhou episode link, a YouTube video, or a local audio file.

podwise process https://app.podwise.ai/dashboard/episodes/7360326
podwise process https://www.xiaoyuzhoufm.com/episode/abc123
podwise process https://www.youtube.com/watch?v=d0-Gn_Bxf8s
podwise process ./episode.mp3

Read what was processedCopy Anchor Link

podwise get summary https://app.podwise.ai/dashboard/episodes/7360326
podwise get transcript https://app.podwise.ai/dashboard/episodes/7360326

Fetching an episode that is already processed does not spend credits. An episode has to be processed first — either by you (web app or podwise process) or by Podwise, which processes new episodes of popular shows automatically.

Output and pipingCopy Anchor Link

Commands print their result to standard output, so you can redirect it to a file or pass it to another program:

podwise get summary <episode-url> > episode.md
podwise search episode "design systems" --limit 5 | head -40

Search commands take --limit; the rest of the flags differ per command, so run podwise --help or podwise <command> --help before adding one to a script.

Local MCP serverCopy Anchor Link

podwise mcp speaks MCP over stdin/stdout. Desktop clients that can launch a command use it instead of the hosted server:

{
  "mcpServers": {
    "podwise": {
      "command": "podwise",
      "args": ["mcp"]
    }
  }
}

The local server uses the CLI's saved token, so there is no OAuth step. Its tools cover discovery, processing, summaries, transcripts, Q&A, chapters, mind maps, highlights, keywords, Ask, follow and history — a smaller set than the hosted server's 38 tools. The hosted and local comparison has the trade-offs.

Agent skillsCopy Anchor Link

The CLI ships with agent skills — named workflows such as catch-up, weekly recap, episode notes, and topic research — that teach an agent which commands to run:

npx skills add hardhackerlabs/podwise-cli

The CLI must already be installed and signed in; the skill calls it. Skills & Plugins explains what each workflow does and how the plugin packaging differs.

RelatedCopy Anchor Link

• MCP Server: the hosted server, its 38 tools, and per-client setup.
• AI Agent: which of the three routes fits your setup.
• Authentication: tokens, OAuth, and what to do on 401.
• podwise-cli on GitHub: source, releases, and the full command reference.