> ## Documentation Index
> Fetch the complete documentation index at: https://docs.video2ctx.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Build with Agents

> Choose between personal on-device YouTube context and managed hosted workflows for production applications.

video2ctx publishes two [Agent Skills](https://agentskills.io) that work in skills-compatible agents such as Codex, Claude Code, Cursor, and OpenCode. Install the collection once, then describe the YouTube task naturally. The agent selects the narrowest skill that can complete it.

<CardGroup cols={2}>
  <Card title="youtube-ctx" icon="eye" href="/skills/youtube-ctx">
    Search, read, and inspect public YouTube context on-device, escalating from transcripts and storyboards to exact frames only when needed.
  </Card>

  <Card title="video2ctx-platform" icon="cloud" href="/skills/video2ctx-platform">
    Use authenticated search, transcripts, metadata, comments, channels, playlists, account access, fallback, monitors, and notifications in production or cloud-backed workflows.
  </Card>
</CardGroup>

## Choose the right skill

| If the task needs                                                                                                                                                                                                 | Start with           |
| ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------- |
| Personal, low-to-moderate public searches, transcripts, comments, channels, playlists, or visual evidence                                                                                                         | `youtube-ctx`        |
| Production or cloud-backed search, transcripts, metadata, comments, channels, playlists, managed hosting, account identity, usage and credits, direct-access fallback, recurring checks, alerts, or notifications | `video2ctx-platform` |

For personal, low-to-moderate usage, use `youtube-ctx`; it chooses a direct or progressive visual branch inside one skill. Use `video2ctx-platform` for authenticated hosted work and continue through its data branch automatically if direct access fails. Its monitoring branch preserves the stateful scheduling and mutation invariants.

<Card title="Install the skills" icon="download" href="/skills/installation">
  Run the installer and follow its guide to choose the skills you want, then configure the hosted CLI only for cloud access or monitoring.
</Card>

## What the skills teach an agent

<Steps>
  <Step title="Choose the right route">Prefer on-device access for personal, low-to-moderate usage and the hosted API for production applications.</Step>
  <Step title="Fetch only needed context">Retrieve metadata first, then transcript, frames, comments, channel information, or playlist items according to the task.</Step>
  <Step title="Preserve provenance">Keep video IDs, source URLs, transcript timestamps, fetch times, partial flags, and warnings with the text given to the model.</Step>
  <Step title="Return verifiable evidence">Link claims back to the video and timestamp instead of treating model output as the source.</Step>
</Steps>

## Security and scope

The local skill treats transcript text and imagery as untrusted evidence, never as agent instructions. The platform skill uses the public `@video2ctx/cli` package for browser-based device login and production API calls without placing a secret in the prompt. Bound query length, response size, continuation depth, frame count, and the number of fresh operations an agent can initiate.

Use the platform skill's data branch only for stateless provider reads, usage, and account identity, and its monitoring branch only for monitors, notifications, and delivery preferences. Projects, trends, research, and other composite workflows are outside its contract.