Start free

Platforms MCP server

MCP server · 4 tools · 3 endpoints

Twitch MCP server

Streamer profiles, channel video listings and clips from Twitch, each a 2-credit call for your agent.

Get a free key Connect in a minute

100 credits on signup · No card · Sign in with OAuth or paste a key

scrapersocial-twitch mcp.scrapersocial.com/twitchconnected
you

Get the Twitch profile for <channel handle>: followers and bio

agent
calls get_profile 2 credits / profile
{ "handle": "<handle>" }
server
returns
followers number subscribers number bio string
Shape only. Keys follow the tool's summary; a real call returns the full record. 1 call · 2 credits

Connect

One URL. Your client already knows what to do with it.

The first call opens a sign-in. New accounts get 100 credits, no card. Every client below shares one account and one balance.

Get a free key
server url https://mcp.scrapersocial.com/twitch
claude.aiOAuth

Settings → Connectors → Add custom connector, then paste the server URL.

mcp.scrapersocial.com/twitch

Web, desktop and mobile. Nothing else to paste.

Claude Codeone command

Run this in a terminal.

$ claude mcp add --transport http scrapersocial-twitch https://mcp.scrapersocial.com/twitch

Approve the sign-in prompt on first use.

Cursorone click

Installs the server and opens the MCP panel.

Add to Cursor
or paste the JSON into .cursor/mcp.json
{ "mcpServers": { "scrapersocial-twitch": { "url": "https://mcp.scrapersocial.com/twitch", "headers": { "Authorization": "Bearer sk_live_..." } } } }
Other clients · VS Code, Windsurf, Gemini CLI, anything that speaks MCP
VS Codeone click or JSON
Add to VS Codeor run MCP: Add Server, choose HTTP and use:
{ "name": "scrapersocial-twitch", "type": "http", "url": "https://mcp.scrapersocial.com/twitch" }
Windsurf~/.codeium/windsurf/mcp_config.json
{ "mcpServers": { "scrapersocial-twitch": { "serverUrl": "https://mcp.scrapersocial.com/twitch", "headers": { "Authorization": "Bearer sk_live_..." } } } }
Gemini CLI~/.gemini/settings.json
{ "mcpServers": { "scrapersocial-twitch": { "httpUrl": "https://mcp.scrapersocial.com/twitch", "headers": { "Authorization": "Bearer sk_live_..." } } } }
Any clientStreamable HTTP · OAuth 2.1 or Bearer
$ npx add-mcp https://mcp.scrapersocial.com/twitch

One account and one balance across every client and the every-platform hub at mcp.scrapersocial.com.

What it does

What your agent can do with Twitch

Creator research on Twitch usually starts with three questions: who is this streamer, what have they published lately, and what’s in this clip someone sent me. ScraperSocial’s Twitch spoke answers each with one call.

get_profile takes a channel handle and returns the public profile record: follower count, bio and the channel’s totals. get_channel_posts takes the same handle and lists the channel’s recent videos with per-video metrics, paged. The third read, a single clip, isn’t wrapped as a tool; your agent calls call_endpoint with path: /v1/twitch/clip and either the clip URL or its slug. Run list_endpoints when you want the inputs and prices of all three endpoints laid out before you spend anything.

Twitch is a streaming platform, and the server reads it like any public site: what a logged-out visitor sees on a channel page, a video list or a clip page, returned as JSON. You don’t register a Twitch application or manage tokens.

Each call costs 2 credits, and the video list is 2 credits per video returned. Connect https://mcp.scrapersocial.com/twitch and ask about a channel by name.

Four jobs it finishes on its own

01

Vetting a streamer for sponsorship

Ask for the profile and the last 50 videos. The agent reports follower count, how often the channel publishes and how views are distributed, which is most of what a brand needs before a first conversation.

02

Cross-platform creator profiles

Combine `get_profile` here with the equivalent tools on the other spokes and the agent builds one record per creator: Twitch followers next to TikTok and Instagram figures, from a single prompt. Handles differ per network, so give it each one.

03

Clip triage

Paste a clip link and ask what it is. `/v1/twitch/clip` returns the clip's public details so the agent can tell you the channel and the title without opening a browser.

04

Tracking a watchlist

Keep a list of handles and have the agent call `get_channel_posts` on each weekly. It can flag channels that went quiet, or ones whose latest videos jumped in views.

Limits · read before you connect

What this does not do

Written by a person, not generated. If something here is wrong, tell us.

The server reads public pages only. Live status, viewer counts for a stream in progress and chat aren’t covered; there’s no real-time endpoint, and cached responses make ‘is this channel live right now’ unreliable here. Use Twitch’s own API for that. Subscriber revenue, dashboard analytics and anything behind a login are out of scope. A channel’s stream schedule isn’t offered. Profile and clip calls are 2 credits each; the video list is 2 credits per video, fifty per call with a cursor, and a channel with no videos still costs 1. Pass fresh: true to bypass the cache at the same price.

Tools

4 tools, same names as the every-platform hub

Two do one Twitch thing each. The last two reach all 3 endpoints by path, so nothing in the catalog is out of an agent's reach.

Full credit table →

Get an account profile

get_profile

Profile, page or channel record for a public account: follower and subscriber counts, bio, and totals.

2 credits / profile

List an account's posts

get_channel_posts

Recent public posts or videos from one account, with per-item metrics.

2 credits / video
Reaches the other 1 endpoint

List every Twitch endpoint

list_endpoints

Browse every Twitch endpoint, its input and its credit cost. Free, and calls nothing upstream.

free

Call any Twitch endpoint by path

call_endpoint

Run any Twitch endpoint by its path, so nothing Twitch offers is out of reach of an agent.

endpoint's own cost

Pricing

100 free credits is 50 profile reads.

Then $5 a month for 1,000 credits, or $4 per 1,000 as a top-up. No card to start. Every call costs 2 credits.

50 profile readsget_profile · 2 each
Get a free key 100 credits on signup · $5 for 1,000 · full credit table

Prompts

Things to ask once it is connected

Copy one, swap the placeholder, paste it into your client.

Get the Twitch profile for <channel handle>: followers and bio

List the last 50 videos from <channel handle> with their view counts

What is this clip? <twitch clip url>

Compare <channel handle 1> and <channel handle 2> on followers and how often they post videos

Check these channels for new videos this week: <handle 1>, <handle 2>, <handle 3>

Questions

Before you connect

What does a profile lookup cost? #

2 credits. Video lists are 2 credits per video, so 50 videos are 100 credits, which happens to be the free allowance a new account gets. Beyond that, $5 is 1,000 credits.

Can the agent tell me whether a streamer is live? #

Not reliably. There's no live-status endpoint, and responses may come from cache. Twitch's own API is the right tool for real-time state.

Is ScraperSocial affiliated with Twitch or Amazon? #

No. It's an independent service, not an official Twitch API, and has no connection to Twitch or its parent company. It reads public pages.

What's the handle? #

The channel name from the URL, without the @. For clips, `/v1/twitch/clip` accepts either the full clip URL or the clip slug.

Can I get a channel's schedule? #

Not right now. There's no schedule endpoint on this spoke. `list_endpoints` always shows the live set, so check there if that changes.