AI assistants, MCP

Use YouTube and Instagram data inside ChatGPT, Claude and Cursor

Both Actors are available as tools through Apify's hosted MCP server. Add one URL to your AI client, sign in with your Apify account, and ask for YouTube or Instagram data in plain English. The assistant runs the Actor, reads the rows and answers. Each run bills your own Apify account per result.

• YouTube $2 / 1k, Instagram $1.90 / 1k
• One URL per client
• OAuth sign-in, no token in the URL
• ChatGPT, Claude, Cursor, VS Code

Pick a server URL

Each URL points at Apify's hosted MCP server and exposes only the listed tools: the Actor itself plus fetch-actor-details, which lets the assistant read the Actor's README, input options and pricing before its first paid run. There is no token in the URL. Your client signs in to Apify with OAuth the first time you connect.

YouTube scraper as an MCP tool

With the YouTube URL connected, the assistant sees one tool for the YouTube Scraper and one that reads its README. It fills the Actor's input from your question: search terms, a channel name, video links, how many comments, whether to add dislike estimates. You see the input it sends, and it costs what the same run costs from the Console: $2 per 1,000 videos, comments included.

ChatGPT

1. Open ChatGPT on the web and go to Settings, Apps & Connectors, Advanced settings.
2. Turn on Developer mode. OpenAI offers custom connectors on some plans only, so check yours if you don't see the option.
3. Choose Create, paste one of the URLs above and set Authentication to OAuth.
4. Sign in to Apify when prompted, then enable the connector in a chat.

Claude

Claude.ai and Claude Desktop: Settings, Connectors, Add custom connector, paste the URL, press Connect and sign in to Apify.

Claude Code: run this in your terminal.

claude mcp add --transport http yugenox-social "https://mcp.apify.com/?tools=fetch-actor-details,yugenox/instagram-scraper,yugenox/youtube-scraper"

claude mcp add --transport http yugenox-youtube "https://mcp.apify.com/?tools=fetch-actor-details,yugenox/youtube-scraper"

claude mcp add --transport http yugenox-instagram "https://mcp.apify.com/?tools=fetch-actor-details,yugenox/instagram-scraper"

Instagram scraper in Claude Code

After claude mcp add, start Claude Code and run /mcp to sign in to Apify. Then ask in your terminal, for example "get the last 30 reels from @nasa with transcripts and save them to reels.csv": Claude runs the Instagram Scraper, reads the rows and writes the file. It is a quick way to pull data into a script or notebook you're already working on.

Cursor and VS Code

These buttons open your editor and pre-fill the server. You'll be asked to sign in to Apify the first time a tool runs.

Any other MCP client that supports remote HTTP servers can use this config:

{
  "mcpServers": {
    "yugenox-social": {
      "type": "http",
      "url": "https://mcp.apify.com/?tools=fetch-actor-details,yugenox/instagram-scraper,yugenox/youtube-scraper"
    }
  }
}

Things to ask

• "Get the last 50 videos from @Fireship with likes and estimated dislikes, and tell me which topics got the best like-to-dislike ratio."
• "Pull the top 20 comments from each of these 10 videos and summarise the complaints."
• "Get the last 30 reels from @nasa with transcripts and tell me which hooks they open with."
• "Find 20 accounts similar to [a creator you already work with] with more than 10,000 followers and an email in their bio."
• "What are people posting at this location this week?" followed by a location link.

Dislike figures are estimates from Return YouTube Dislike, and the assistant should present them that way.

Keeping costs predictable

• Say how many. "Max 50 videos" or "the last 30 reels" keeps each run small. For reference, 50 videos with comments cost $0.10 and 30 thirty-second reels with transcripts about $0.18.
• Cap each run. Pay-per-event Actors accept a maximum cost per run (maxTotalChargeUsd in the API), and Apify lets you set usage limits on your account.
• Nothing runs on its own. The assistant only runs the Actor when you ask it something that needs data.

Security notes

• The URLs carry no credentials. OAuth ties each connection to the Apify account that signed in.
• For clients without OAuth, send your token as a header: Authorization: Bearer <APIFY_TOKEN>. Never put a token in the URL or share a config file that contains one.
• The server exposes only the tools in the URL, so the assistant can't run other Actors through it.

Using an agent framework without MCP

If your agent calls functions directly (LangChain, the OpenAI Agents SDK, your own loop), wrap the Actor call as a tool. This one returns 20 search results with 10 top comments each, for $0.04:

// Save as run.mjs, then: npm i apify-client && APIFY_TOKEN=... node run.mjs
import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: process.env.APIFY_TOKEN });

const run = await client.actor('yugenox/youtube-scraper').call({
  searchTerms: [
    "ai agents tutorial"
  ],
  maxItems: 20,
  maxComments: 10
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
for (const item of items) console.log(item.title, item.viewCount, item.commentsCount);

import os
from apify_client import ApifyClient  # pip install apify-client

client = ApifyClient(os.environ["APIFY_TOKEN"])

run = client.actor("yugenox/youtube-scraper").call(run_input={
    "searchTerms": ["ai agents tutorial"],
    "maxItems": 20,
    "maxComments": 10,
})
if run is None:
    raise RuntimeError("Run did not finish")

for item in client.dataset(run.default_dataset_id).iterate_items():
    print(item.get("title"), item.get("viewCount"), item.get("commentsCount"))

# Small jobs: runs the Actor and returns the rows in one call (up to 5 minutes).
curl -X POST "https://api.apify.com/v2/acts/yugenox~youtube-scraper/run-sync-get-dataset-items" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"searchTerms":["ai agents tutorial"],"maxItems":20,"maxComments":10}'

Last updated 2026-10-06.