> ## Documentation Index
> Fetch the complete documentation index at: https://api-tools.memories.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Video Searching Agent

> Official OpenClaw skill on ClawHub powered by the managed Memories.ai Video Searching API.

This page documents the current **OpenClaw skill on ClawHub** for video searching and analysis.
The hosted skill is backed by the **managed Memories.ai Video Searching API**, so the primary path no longer requires cloning or running the older local runtime.

## OpenClaw Skill on ClawHub

<CardGroup cols={3}>
  <Card title="Launch Skill on ClawHub" icon="rocket" href="https://clawhub.ai/memories-ai-official/video-searching">
    Open the hosted workflow and run `/video_search` queries directly in ClawHub.
  </Card>

  <Card title="API Reference" icon="bolt" href="/visual-agents/video-searching-api">
    See the managed API endpoint, SSE behavior, and request examples that power this skill.
  </Card>

  <Card title="Skill Identity" icon="code">
    Type: OpenClaw Skill

    <br />

    Host: ClawHub

    <br />

    Slug: `memories-ai-official/video-searching`

    <br />

    Current published version: `1.0.1`
  </Card>
</CardGroup>

<CardGroup cols={2}>
  <Card title="Starter Prompt" icon="wand-magic-sparkles">
    Try: `/video_search Find 10 high-engagement short videos about AI coding tools from the past week, and summarize hook patterns.`
  </Card>

  <Card title="Runtime Requirements" icon="wrench">
    Requires `MEMORIES_API_KEY` plus standard `curl` and `jq` support in the OpenClaw environment.
  </Card>
</CardGroup>

<Note>
  Official skill page:
  [clawhub.ai/memories-ai-official/video-searching](https://clawhub.ai/memories-ai-official/video-searching)
</Note>

## What Changed In The Current Skill

* The hosted ClawHub skill now runs on the managed Memories.ai Video Searching API instead of the older self-hosted Python runtime.
* The user-facing invocation command is `/video_search`.
* Authentication is simplified around `MEMORIES_API_KEY`.
* The hosted flow removes local repo cloning and dependency bootstrapping from the primary path.
* Results are streamed and returned as structured video references with summary text.

## What It Can Do

<CardGroup cols={2}>
  <Card title="Multi-Platform Search" icon="magnifying-glass">
    Searches YouTube, TikTok, Instagram, and X/Twitter from a single prompt.
  </Card>

  <Card title="Structured Video References" icon="file-lines">
    Returns answer text, source links, relevance notes, and key metrics when available.
  </Card>

  <Card title="Recency Controls" icon="sliders">
    Supports time filters such as `past_24h`, `past_week`, `past_month`, and `past_year`.
  </Card>

  <Card title="Clarification Flow" icon="comments">
    Can ask follow-up questions when a request is too vague to search reliably.
  </Card>
</CardGroup>

## Invocation Pattern

```text theme={null}
/video_search Find top SaaS UGC videos on TikTok from the past week
```

Optional controls in the underlying API include platform filters, result limits, agent step caps, and time-frame constraints.
For raw request and response examples, see [Video Searching API](/visual-agents/video-searching-api).

## Best Fit Scenarios

* Industry and topic trend discovery
* Brand and product video analysis
* Creator scouting and comparison
* Cross-platform benchmarking
* Research workflows that need traceable video references

## Legacy Runtime Context

Earlier versions of this page described the open-source local runtime in more detail.
That repository is still useful if you want to inspect or adapt the older self-hosted workflow, but it is no longer the primary path for the current ClawHub skill.

<CardGroup cols={2}>
  <Card title="Managed API Docs" icon="book" href="/visual-agents/video-searching-api">
    Use this if you want the same hosted capabilities outside ClawHub via SSE.
  </Card>

  <Card title="Legacy Runtime Repository" icon="code" href="https://github.com/Memories-ai-labs/video-searching-agent">
    Reference implementation for the earlier local runtime flow.
  </Card>
</CardGroup>

<Note>
  The current ClawHub skill metadata advertises macOS/Linux support and requires `MEMORIES_API_KEY`, `curl`, and `jq`.
</Note>