What is Watch Video (YouTube)?
Watch Video (YouTube) is a ai & search tool. With OSP.net, your AI agent connects to it directly so it can act on your behalf instead of just answering questions about it.
What your OSP agent can do with Watch Video (YouTube)
- Read a YouTube video's transcript (captions)
- Watch the video frame-by-frame with vision — describe what's shown
- Answer questions and summarize from both what's seen and heard
How to connect Watch Video (YouTube)
- No key needed to start: paste a YouTube link and ask your agent to watch it — it reads the captions out of the box. The keys below unlock the extras.
- For frame-by-frame VISION (watching what's on screen, not just the captions), add an OpenRouter key — open the Keys page, Create Key, copy it (sk-or-…), and add a little credit (vision is pay-as-you-go). openrouter.ai → Keys
- Optional — for videos with NO captions, add a Whisper transcription key so your agent can still read what's said. Easiest is a free Groq key (gsk_…). console.groq.com → API Keys
- Or use an OpenAI key (sk-…) for Whisper transcription instead of Groq. platform.openai.com → API keys
- Paste whichever keys you want here and Connect. The Watch Video skill picks them up automatically the next time your agent runs it.
Security: Captions-only watching needs NO key. The OpenRouter key only spends your OpenRouter credits (frame vision); the Groq/OpenAI keys are used only to transcribe video audio. Each is stored encrypted in Vault and injected only at your agent's runtime — never written to disk or logs. Rotate any of them any time from the provider's dashboard.
Related ai & search integrations
Frequently asked questions
- Can OSP.net connect to Watch Video (YouTube)?
- Yes. Watch Video (YouTube) is a native OSP.net integration — you bring your own Watch Video (YouTube) key or token, paste it in your dashboard, and your agent restarts live.
- What can my OSP agent do with Watch Video (YouTube)?
- Give your agent a YouTube URL and it reads the transcript AND watches the video frame-by-frame — answering questions about what's actually on screen, not just what's said. Specifically: Read a YouTube video's transcript (captions); Watch the video frame-by-frame with vision — describe what's shown; Answer questions and summarize from both what's seen and heard.
- Is Watch Video (YouTube) a native integration or via the gateway?
- Watch Video (YouTube) is a native, baked-in integration. You connect it with your own credentials, which are stored in an encrypted vault and injected only at runtime.
- Is my Watch Video (YouTube) data secure with OSP.net?
- Yes. Your Watch Video (YouTube) credentials live in an encrypted secrets vault, are injected only at container runtime, and are never written to disk in plaintext or used to train any model. Each customer runs in a fully isolated instance.