How to Build a Daily YouTube Digest Agent with OpenClaw: Never Miss a Video Again

AI YouTube Summarization: The Problem This Agent Solves

You follow 20 technical YouTube channels. Fireship, Andrej Karpathy, ThePrimeagen, AI Explained, and sixteen others. New videos appear irregularly — sometimes daily, sometimes not for two weeks. You have no reliable system for catching them.

You open YouTube when you remember, scroll through the homepage algorithm that surfaces the wrong things, and occasionally realize three days later that a channel you care about released something relevant to a decision you already made. The weekend catch-up session — 90 minutes of backlogged content you half-watch at 2× speed — is not a system. It's a failure mode with a coping strategy.

The fix: configure OpenClaw to check each channel daily for new uploads, pull the transcript or description, summarize each video to a 2–3 sentence technical brief with a Watch/Skim/Skip recommendation, and deliver the full digest before 8 AM. Twenty channels, zero manual checking, consistent daily coverage. This tutorial — part of the OpenClaw Social Media Use Cases category — builds exactly that agent.

By the end of this tutorial you'll have a YouTube digest agent running via WisGate — monitoring your channel list, ingesting transcripts or descriptions, and delivering a ranked daily summary before your morning starts. Test the summarization output against a real transcript at wisgate.ai/studio/image before scheduling the first run. Get your API key at wisgate.ai/hall/tokens.

What the Daily YouTube Digest Agent Does

Inputs:

• Channel list: up to 20 YouTube channel handles or IDs
• Per-channel preference tags: topics you prioritize per channel (optional)
• Time window: videos published in the last 24 hours

Processing:

1. Query the YouTube Data API for new uploads per channel (free tier quota is sufficient for personal use at 20 channels)
2. For each new video: fetch the transcript if available; fall back to description
3. Pass each video's transcript or description to Claude via WisGate for summarization
4. Return a structured brief per video with a Watch/Skim/Skip recommendation

Output: structured daily digest — one entry per new video — delivered to Slack, email, or a local file via cron at a configured time.

Components needed: OpenClaw configured with WisGate, a YouTube Data API v3 key, and the YAML and API call in this tutorial. For a simpler starting point with no transcript infrastructure, see the Daily Reddit Digest tutorial — the same base configuration applies.

LLM Video Transcript Summary: Choosing Between Haiku and Sonnet

This is the model decision that doesn't arise in simpler digest automations. Transcript length changes the correct tier — using Haiku on a 30-minute technical lecture produces a shallow summary of the opening minutes, not the full content.

The decision rule:

Description-only mode is the correct default for most channels. YouTube descriptions average 200–400 tokens — well within Haiku's sweet spot for fixed-schema summarization at low cost. No transcript fetching infrastructure required.

Transcript mode is for channels where summary depth matters. For a channel like Andrej Karpathy — where the value is in two hours of technical depth, not the title — Haiku is the wrong choice regardless of cost per call. Switch to claude-sonnet-4-5 when passing full transcripts for medium and long-form videos. Confirm both model prices from https://wisgate.ai/models before finalizing your channel configuration.

OpenClaw API YouTube Automation: WisGate Configuration

Step 1 — Locate and Open the Configuration File

OpenClaw stores its configuration in a JSON file in your home directory. Open your terminal and edit the file at:

Using nano:

nano ~/.openclaw/openclaw.json

Step 2 — Add the WisGate Provider to Your Models Section

Copy and paste the following into the models section of your openclaw.json. This example registers both Haiku and Sonnet so you can assign the correct model per channel:

"models": {
  "mode": "merge",
  "providers": {
    "moonshot": {
      "baseUrl": "https://api.wisgate.ai/v1",
      "apiKey": "YOUR-WISGATE-API-KEY",
      "api": "openai-completions",
      "models": [
        {
          "id": "claude-haiku-4-5-20251001",
          "name": "Claude Haiku 4.5",
          "reasoning": false,
          "input": ["text"],
          "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
          "contextWindow": 256000,
          "maxTokens": 8192
        },
        {
          "id": "claude-sonnet-4-5",
          "name": "Claude Sonnet 4.5",
          "reasoning": false,
          "input": ["text"],
          "cost": { "input": 0, "output": 0, "cacheRead": 0, "cacheWrite": 0 },
          "contextWindow": 256000,
          "maxTokens": 8192
        }
      ]
    }
  }
}

Replace YOUR-WISGATE-API-KEY with your key from wisgate.ai/hall/tokens. Full model list and confirmed pricing: https://wisgate.ai/models

Step 3 — Save, Exit, and Restart OpenClaw

1. Press Ctrl + O → Enter to save
2. Press Ctrl + X to exit nano
3. Press Ctrl + C to stop the current session, then run:

openclaw tui

Note: OpenClaw was previously known as ClawdBot and MoltBot — these steps apply to all versions.

OpenClaw Use Cases: Annual Cost for a 20-Channel YouTube Digest Agent

Assumptions:

• 20 channels, average 1 new video/channel/day = 20 videos/day
• 14 channels in description mode (Haiku), 6 channels in transcript mode (Sonnet)
• Description input: ~300 tokens avg; transcript input: ~5,000 tokens avg
• Output per video: ~200 tokens

Annual token volume:

Cost comparison — confirm all per-token pricing from https://wisgate.ai/models before publishing; insert confirmed figures below:

Once confirmed, state the annual saving as a specific dollar figure. At 20 channels and 365 days, the Sonnet transcript calls are the dominant cost driver — calculate those first, then add Haiku for the description-mode channels.

OpenClaw Use Cases: 20 Channels, Zero Manual Checking

Copy the YAML. Add your channel list. Set mode: transcript and model_override: claude-sonnet-4-5 for your high-value long-form channels. Validate one summary in AI Studio. Activate the schedule.

The first digest runs tomorrow morning. Every new video from every channel you follow — summarized, ranked, and delivered before the workday starts.

Your channel list is the only thing standing between you and tomorrow's first digest. Get your WisGate key at wisgate.ai/hall/tokens — trial credits included, no commitment. Copy the YAML from above, add your channels, then paste one real video description into wisgate.ai/studio/image and verify the summary quality before activating the schedule. Your single next action: generate the API key.

All per-token pricing requires confirmation from wisgate.ai/models before publishing the cost comparison table. Insert confirmed Haiku and Sonnet rates and calculate the annual delta before this article goes live.