___  ___  ___/ / _/ /__ _    __
   / _ \/ _ \/ _  / _/ / _ \ |/|/ /
  / .__/\___/\_,_/_//_/\___/__,__/
 /_/

Your podcast listening, working for you. Extract guests, ideas, and insights from your Apple Podcasts library using AI.

Quick Start

npx podflow init                          # create config
npx podflow digest --dry-run              # preview episodes
npx podflow digest --max-episodes 10      # process 10 episodes

Or install globally:

npm install -g podflow
podflow digest

Commands

| Command | Description | | ------- | ----------- | | podflow init | Create config at ~/.podflow/ | | podflow digest | Process new episodes (default) | | podflow digest --dry-run | Preview without API calls | | podflow digest --backfill | Process all episodes | | podflow subs | List podcast subscriptions | | podflow stats | Cache statistics | | podflow schedule | Weekly automatic digests + notification (macOS) | | podflow mcp | Start the MCP server (stdio) — ears for agents |

Set it and forget it

podflow init now walks you through setup interactively (provider, key, interests) — no JSON editing. Then:

podflow schedule            # weekly, Mondays 08:00
podflow schedule --daily --time 07:30
podflow schedule --status   # check it
podflow schedule --off      # remove it

Each scheduled run processes new episodes and ends with a macOS notification, so the digest reaches you instead of waiting in a file. Your API key is stored in ~/.podflow/config.json (permissions 0600) so scheduled runs work without shell environment variables.

Agents can use it too (MCP)

podflow ships an MCP server, so AI agents can query your podcast intelligence directly — what was said, by whom, and why it matters.

{
  "mcpServers": {
    "podflow": {
      "command": "npx",
      "args": ["-y", "podflow", "mcp"]
    }
  }
}

Six tools:

| Tool | What an agent gets | | ---- | ------------------ | | search_episodes | Keyword search across titles, guests, ideas, and notes | | get_guest | Cross-episode guest profile: role, company, appearances, why follow | | get_ideas | Extracted ideas filtered by topic and relevance | | list_follow_worthy | Guests flagged worth following, with context | | get_stats | Cache totals and cost | | ingest_feed | Point it at any podcast RSS feed to extract new episodes on demand |

Intelligence is episode-level (descriptions plus Apple transcript snippets where available). Uses your existing ~/.podflow config and provider key; ingest_feed costs pennies per episode, same as digest.

Why podflow?

• Local-first. Reads the Apple Podcasts SQLite database on your Mac. No cloud sync, no account needed.
• Your AI key, your choice. Supports Anthropic, OpenAI, Google, and Ollama (local, free).
• Incremental. Only processes new episodes. Run it weekly, costs pennies.
• Configurable interests. Scores everything against what matters to you.

How It Works

1. Reads the Apple Podcasts SQLite database
2. Prioritises episodes by podcast tier (you configure which podcasts matter)
3. Batches episodes and sends to your chosen AI provider
4. Extracts guests, ideas, people mentioned, and relevance scores
5. Caches results incrementally (interrupted runs resume)
6. Generates a markdown digest

Configuration

After podflow init, edit ~/.podflow/config.json:

{
  "about": "A founder building developer tools",
  "interests": [
    {
      "name": "Customer Acquisition",
      "keywords": ["pricing", "growth", "retention", "PLG"],
      "why": "Growing from 0 to first paying customers"
    },
    {
      "name": "AI & Agents",
      "keywords": ["AI", "LLM", "agent", "automation"],
      "why": "Building AI-powered features"
    }
  ],
  "provider": "anthropic",
  "model": "claude-haiku-4-5-20251001",
  "outputPath": "./podflow-digest.md"
}

Podcast Tiers

Edit ~/.podflow/podcasts.json to prioritise specific podcasts:

{
  "podcasts": {
    "My Favourite Podcast": { "tier": 1, "extractGuests": true, "extractIdeas": true },
    "Entertainment Only": { "tier": 4, "extractGuests": false, "extractIdeas": false }
  },
  "defaults": { "tier": 3, "extractGuests": true, "extractIdeas": true }
}

Tiers 1-3 are processed by default. Tier 4 is skipped (use --include-all to override).

Global Flags

--dry-run               Preview without API calls
--backfill              Process all episodes
--max-episodes <n>      Limit episodes (default: 50)
--tier <n>              Only process specific tier (1-4)
--include-all           Include tier 4
--provider <name>       anthropic, openai, google, ollama
--model <name>          Override model name
--output <path>         Output file path
--recent <days>         Days of history (default: 365)
--verbose               Show episode details
-q, --quiet             Suppress output except errors

AI Providers

| Provider | Env Var | Cost / 1,000 eps | Best For | | -------- | ------- | ----------------- | -------- | | Anthropic (Haiku) | ANTHROPIC_API_KEY | ~$2 | Best quality/cost ratio | | OpenAI (GPT-4o-mini) | OPENAI_API_KEY | ~$0.50 | Cheapest cloud option | | Google (Gemini Flash) | GOOGLE_GENERATIVE_AI_API_KEY | ~$0.25 | Budget option | | Ollama (local) | None | Free | Privacy, no API key |

podflow digest --provider openai --model gpt-4o-mini
podflow digest --provider ollama --model llama3.2

Output

Generates a markdown digest with:

• People to Follow -- guests flagged as worth following, with roles, companies, and social links
• Ideas by Topic -- actionable insights grouped by your configured interest areas
• High Relevance -- episodes scored 7+ for relevance to your work
• Recent Insights -- last 30 days of extracted intelligence
• Guest Index -- every guest across all episodes, cross-referenced

Cost

| Scenario | Episodes | Cost (Haiku) | | -------- | -------- | ------------ | | Weekly run | ~20 | ~$0.03 | | Monthly | ~80 | ~$0.12 | | Full backfill | 1,000 | ~$2.00 |

Requirements

• macOS with Apple Podcasts (Sync Library enabled)
• Node.js 20+
• An API key for your chosen provider (or Ollama for local)

Contributing

See CONTRIBUTING.md for dev setup and guidelines.

Changelog

See CHANGELOG.md for release history.

Part of Total Audio

Tools I build for music PR, by Chris Schofield. Part of Total Audio Promo.

| Project | Description | |---------|-------------| | TAP | Campaign management for music PR agencies | | totalaud.io | Release planning for emerging artists | | SpotCheck | Spotify playlist validation | | Newsjack | Music industry newsjacking | | Podflow | Podcast intelligence for music PR | | Sink | Contact data hygiene CLI |

Questions? Reach me on X/@chrisschouk or [email protected].

Licence

MIT