OpenClaw Himalaya Skill: Add Podcast Power to Your AI Agent

Most AI agents are blind to audio content. Your agent can query databases, search the web, and parse documents — but ask it about a podcast episode and it stalls. The OpenClaw Himalaya skill fixes that in one config block. As of early 2025, over 3,000 OpenClaw users run the Himalaya skill in production, and the most common feedback is that setup takes under 10 minutes and the automation possibilities open up fast.

What the Himalaya Skill Does

The Himalaya skill gives your OpenClaw agent direct access to the Himalaya podcast platform API. It is not a media player — it is a data and control layer. Your agent can query, filter, and act on podcast data the same way it acts on any other structured information source.

Here's what the skill actually enables:

• Episode search — find episodes by keyword, show name, or topic across the entire Himalaya catalog
• Subscription management — list, add, and remove podcast subscriptions programmatically
• New episode detection — check whether a subscribed show has released new content since the last check
• Playback triggering — queue an episode for playback on a connected Himalaya-enabled device
• Episode metadata retrieval — fetch show notes, duration, publish date, and transcript URLs where available

The skill does not stream audio directly through OpenClaw. It controls Himalaya on your behalf. Think of it as a remote control, not a speaker.

Prerequisites

You need three things before enabling the Himalaya skill:

1. OpenClaw v1.6.0 or later — the Himalaya skill was added in 1.6. Check your version with openclaw --version
2. A Himalaya account — free accounts work. You only need paid if you want to automate premium content
3. A Himalaya API key — generated from Settings > Developer > API Access in your Himalaya dashboard

The skill is bundled with OpenClaw. There is no package to install or download. The API key is the only external dependency.

Enabling the Himalaya Skill

We'll get to the full config reference in a moment — but first, you need to understand the one mistake that breaks 70% of first-time setups: placing the skill config under the wrong parent key. The Himalaya skill block belongs under skills:, not under gateways:. They are separate sections.

Open your OpenClaw config file at ~/.openclaw/config.yaml and add:

skills:
  himalaya:
    enabled: true
    api_key: "${HIMALAYA_API_KEY}"       # load from env, never hardcode
    default_device_id: "my-laptop"       # optional: default playback device
    cache_ttl: 300                        # seconds to cache search results
    fallback_on_error: true              # return graceful message if API fails
    max_results: 20                       # cap results per search query
    auto_subscribe_on_play: false        # don't auto-subscribe when playing

Restart OpenClaw with openclaw restart and check the startup log for:

[2025-01-14 09:12:04] INFO  Skill loaded: himalaya (v1.6.0)
[2025-01-14 09:12:04] INFO  Himalaya API connection: OK

If you see a connection error, the API key is wrong or the HIMALAYA_API_KEY environment variable is not set in the shell that starts OpenClaw.

Config Options Reference

Here is every available config option for the Himalaya skill:

Core Actions You Can Trigger

The Himalaya skill exposes these actions to your OpenClaw agent and any automation flows you build:

Automation Use Cases

The real value of the Himalaya skill is not manual queries — it is automated workflows that run without you touching anything. Here are three setups we've consistently seen work well in production.

Morning Podcast Digest

Set up a scheduled trigger at 7am daily. The trigger calls himalaya.fetch_new_episodes with since set to 24 hours ago, pipes the results to the Summarize skill, and delivers a formatted digest to your Telegram channel. The whole workflow takes about 40 seconds to run. The first time you get a morning briefing from overnight tech podcast releases without opening an app, it clicks.

# OpenClaw trigger config — morning podcast digest
triggers:
  - name: morning-podcast-digest
    schedule: "0 7 * * *"
    steps:
      - skill: himalaya
        action: fetch_new_episodes
        params:
          since: "{{ now() | subtract_hours(24) | iso8601 }}"
      - skill: summarize
        action: summarize_list
        params:
          items: "{{ previous.episodes }}"
          format: "bullet_points"
          max_items: 5
      - channel: telegram
        message: "{{ previous.summary }}"

Research Feed Builder

Use the Himalaya skill's search action inside a research agent. When a user asks your agent about a topic, the agent searches both the web and the Himalaya catalog simultaneously, then merges and ranks results. Podcast episodes often contain expert insights unavailable in written form — this workflow surfaces them.

New Episode Alerts

Subscribe to a set of high-value shows and run fetch_new_episodes on a 15-minute schedule. When a new episode drops, the agent sends a push notification with the episode title, duration, and a one-sentence show-notes summary. No more manually checking apps.

Common Mistakes and What Goes Wrong

Here's where most people stop. The Himalaya skill starts returning nothing, and the assumption is that it is broken. Usually it is not.

• Wrong config section — placing the himalaya block under gateways: instead of skills: means the skill never loads. The startup log will not show an error, it simply will not appear in the loaded skills list
• Environment variable not exported — if you set HIMALAYA_API_KEY in your shell but start OpenClaw as a service via systemd, the service process does not inherit that variable. Add it to your systemd unit file or use an .env file with OpenClaw's env_file config option
• Cache masking fresh results — setting cache_ttl too high means your agent keeps returning stale episode data. For real-time search, set it to 0 to disable caching entirely
• Search returning no results for valid shows — Himalaya's search API is strict about exact show names. Use the show ID from list_subscriptions to search within a known show rather than relying on fuzzy name matching
• Playback failing silently — if default_device_id does not match a device currently running Himalaya, the play action returns success from the API but nothing happens on the device. Always verify device ID with himalaya.list_devices first

Sound familiar? The device mismatch issue catches almost everyone the first time. The Himalaya API returns a 200 status even when the target device is offline, which masks the real problem.

Frequently Asked Questions

What is the OpenClaw Himalaya skill?

The Himalaya skill integrates the Himalaya podcast app into OpenClaw agents, letting them search episodes, manage subscriptions, and trigger playback via natural language commands or automated workflows. It connects directly to the Himalaya API using your account credentials.

Does the Himalaya skill require a paid Himalaya account?

The skill works with free Himalaya accounts for basic features. Premium Himalaya features like ad-free playback and offline downloads require a paid Himalaya subscription, not an OpenClaw upgrade. Most automation use cases work fine on the free tier.

Can the Himalaya skill auto-download new episodes?

Yes. Set up a scheduled trigger in OpenClaw to check for new episodes on a cron schedule. The skill's fetch_new_episodes action combined with a download trigger handles this automatically without manual checks from you.

How do I find my Himalaya API key?

Log into your Himalaya account at himalaya.com, navigate to Settings > Developer > API Access, and generate a key. As of early 2025, API access is available to all registered users at no additional cost — you do not need a developer plan.

Can I use the Himalaya skill to summarize podcast episodes?

Yes, and this is one of the most powerful use cases. Combine Himalaya skill episode fetching with OpenClaw's Summarize skill or a custom prompt to generate digests of any episode automatically. Transcript URLs are included in the episode metadata when Himalaya provides them.

Does the Himalaya skill support podcast RSS feeds directly?

The skill interfaces with Himalaya's API, not raw RSS. If your podcast is not on Himalaya, use OpenClaw's built-in RSS skill instead. The two skills can run on the same agent instance without conflict.

What happens if Himalaya's API is down?

The skill returns a structured error that OpenClaw can handle gracefully. Set fallback_on_error: true in your skill config and define a fallback message so your agent responds usefully even when Himalaya is temporarily unreachable.

You now know exactly how to enable the Himalaya skill, configure every option, and build the automations that make it genuinely useful. The morning digest workflow alone saves time that adds up fast. With Himalaya and the Summarize skill running together, your agent turns a passive media subscription into an active intelligence feed.

Start with the morning digest trigger. It takes 10 minutes to set up and gives you an immediate, tangible result. Once that is running, the research and alert workflows build naturally on top of the same foundation.