Copy-pasting YouTube transcripts into Claude is the slow path. The fast path: Scavio's transcript endpoint returns the typed transcript as JSON, then a small Python or JS wrapper streams it into a Claude message. This tutorial shows the wrapper plus a few patterns for summarizing, fact-checking, and follow-up question Q&A.
Prerequisites
- Python 3.10+
- A Scavio API key
- Anthropic API key
Walkthrough
Step 1: Pull a YouTube transcript
Scavio's transcript endpoint returns text plus timing.
Python
import requests, os
API_KEY = os.environ['SCAVIO_API_KEY']
def transcript(video_id):
r = requests.post('https://api.scavio.dev/api/v1/youtube/transcript',
headers={'x-api-key': API_KEY},
json={'video_id': video_id})
return r.json().get('transcript', '')Step 2: Send into Claude
Pass transcript as user-message context.
Python
import anthropic
client = anthropic.Anthropic()
def summarize(video_id):
t = transcript(video_id)
msg = client.messages.create(
model='claude-sonnet-4-6',
max_tokens=512,
messages=[{'role': 'user', 'content': f'Summarize in 5 bullets:\n{t[:8000]}'}])
return msg.content[0].textStep 3: Fact-check claims
Loop claims against Scavio SERP for citations.
Python
def fact_check(claim):
r = requests.post('https://api.scavio.dev/api/v1/google',
headers={'x-api-key': API_KEY},
json={'query': claim, 'num_results': 5})
return r.json().get('organic_results', [])Step 4: Q&A across multiple videos
Concatenate top 3 transcripts, ask Claude.
Python
def multi_video_qa(video_ids, question):
texts = [transcript(vid)[:3000] for vid in video_ids]
context = '\n\n'.join(texts)
msg = client.messages.create(
model='claude-sonnet-4-6',
max_tokens=1024,
messages=[{'role': 'user', 'content': f'{question}\n\n{context}'}])
return msg.content[0].textStep 5: Cache transcripts
Avoid repeat calls on the same video.
Python
from functools import lru_cache
@lru_cache(maxsize=200)
def cached_transcript(vid):
return transcript(vid)Python Example
Python
import os, requests, anthropic
API_KEY = os.environ['SCAVIO_API_KEY']
client = anthropic.Anthropic()
def summarize_yt(vid):
t = requests.post('https://api.scavio.dev/api/v1/youtube/transcript',
headers={'x-api-key': API_KEY}, json={'video_id': vid}).json().get('transcript', '')
msg = client.messages.create(model='claude-sonnet-4-6', max_tokens=512,
messages=[{'role':'user','content':f'Summarize: {t[:8000]}'}])
return msg.content[0].text
print(summarize_yt('dQw4w9WgXcQ'))JavaScript Example
JavaScript
const API_KEY = process.env.SCAVIO_API_KEY;
import Anthropic from '@anthropic-ai/sdk';
const client = new Anthropic();
export async function summarizeYt(vid) {
const r = await fetch('https://api.scavio.dev/api/v1/youtube/transcript', { method: 'POST', headers: { 'x-api-key': API_KEY, 'Content-Type': 'application/json' }, body: JSON.stringify({ video_id: vid }) });
const t = (await r.json()).transcript || '';
const msg = await client.messages.create({ model: 'claude-sonnet-4-6', max_tokens: 512, messages: [{ role: 'user', content: `Summarize: ${t.slice(0, 8000)}` }] });
return msg.content[0].text;
}Expected Output
JSON
Claude returns a 5-bullet summary of the video. Multi-video Q&A produces grounded answers across 3 transcripts. Fact-check returns 5 SERP citations per claim.