That's what MCP changes.
The problem with traditional integrations
Broadcasting and OTT teams spend time switching between tools: the analytics tab, the media library, the ad campaigns module. Each answer requires navigation, filtering, exporting. By the time you have the full picture, the decision moment may have passed.
AI coding assistants and chat tools are already part of many workflows. The missing piece has been a direct, secure bridge between those AI tools and the actual data inside your platform.
MCP (Model Context Protocol) is that bridge.
What is MCP, technically speaking?
MCP is an open protocol — developed by Anthropic and adopted across the AI ecosystem — that allows AI clients to connect to external services through a standardized interface. Instead of copying and pasting data into a chat window, the AI queries your platform directly, in real time, using tools defined by the server.
The protocol runs over standard HTTP + JSON-RPC 2.0, which means any MCP-compatible client can connect without custom code or proprietary adapters.
Mediastream Platform exposes an MCP server at:
https://mcp.platform.mediastre.am
Accounts on the EU region use the EU server instead:
https://eu.mcp.mediastre.am
The server implements protocol version 2024-11-05 — the current stable spec — and is compatible with all major MCP clients including Claude (Anthropic), Cursor, Windsurf, Continue, and any tool built on the open MCP SDK.
Step 1 — Copy the server URL
Go to https://platform.mediastre.am/settings/ai (Settings > MoAI Options), or open the same page on your region's Platform domain. In the MCP Connection section you'll find the server URL for your region. Click the copy icon next to it.
🔒 Don't see the MCP Connection section? It is available to account administrators. If you are one and still don't see it, contact support via the chat button in Mediastream Platform.
Step 2 — Add the server to your AI tool and sign in
Sign in with your Mediastream account. This is the recommended way to connect. You add the server URL to your AI tool and authorize it with your own Platform login. There is no token to copy, paste or rotate, and the connection belongs to the user who authorized it.
Claude (Desktop or web)
Add the server as a custom connector and paste the server URL you copied.
Set Authentication to “required when the server requests it”.
Set OAuth client to Dynamic Client Registration (DCR).
Connect. A Mediastream window opens: sign in with your Platform account.
Review the permissions on the consent screen and click Allow.
⚠️ If Authentication is left as “none”, the connector is added without authentication and every tool answers 401. If that happens, edit the connector and change the two
settings above.
The consent screen lists the Mediastream modules the connection can reach and the permission level: Read only (view your data) or Read and write (view and modify it).
Cursor
Create or edit .cursor/mcp.json at the root of your project. Only the URL is needed. Cursor opens the sign-in window when it connects.
{
"mcpServers": {
"mediastream": {
"url": "https://mcp.platform.mediastre.am"
}
}
}Any other MCP-compatible client
The server follows the standard MCP HTTP transport spec. If your client supports remote MCP servers with OAuth sign-in (Dynamic Client Registration), the pattern is the same: add the server URL and sign in when prompted. Check your client's documentation for the exact steps.
Step 3 (optional) — Use a token instead of signing in
MCP tokens remain available for MCP clients that do not support sign-in, and for programmatic use. On the same page, scroll to MCP Tokens and fill in:
Description: give it a recognizable name (e.g. “Claude Desktop”, “Cursor dev”)
Access: Read, or Read+Write
Expiration: No Expiration, or +1 hour, +1 day, +1 week, +1 month, +1 year
Click Create Token, then use the copy icon in the token list.
⚠️ Treat your token like a password. Anyone with it can reach your account resources. If it is compromised, disable or delete it from the same page and create a new one.
Authentication uses a standard Bearer token header:
Authorization: Bearer YOUR_TOKEN_HERE
For Claude or Cursor, add the header to the server entry:
{
"mcpServers": {
"mediastream": {
"url": "https://mcp.platform.mediastre.am",
"headers": {
"Authorization": "Bearer YOUR_TOKEN_HERE"
}
}
}
}What you can actually do with it
Once connected, your AI client can use the tools of the server. These are the read tools:
Content & Live
getMediaList— Browse your VOD catalog with filters and pagination. Ask “show me all media published in the last 30 days in the Sports category” and get structured results instantly.getMediaById— Pull the full metadata of a specific media item: title, duration, status, categories, encoding details.getLiveStreamList— List your active and past live streams with filtering options.getLiveStreamById— Get full details on a specific live event, including configuration and status.issueAccessToken— Generate access tokens for protected media or live streams programmatically — useful for automated workflows.
Analytics
getMetricsCatalog— Discover all 200+ available metrics and their supported dimensions before querying. Think of it as asking “what can I measure?” before asking “what are the numbers?”queryAnalytics— The most powerful tool. Query streams, duration, unique users, concurrency, bandwidth, engagement, retention, and session behavior — across dimensions like country, device, content ID, category, and distributor.
Advertising
getAdList— Browse your ad campaigns with filters and pagination.getAdById— Pull the full detail of a specific campaign, including configuration and targeting.
With Read and write access, the server also offers tools to create and modify resources, across the modules enabled for your account.
A concrete example
You ask your AI tool: “What were the top 5 most-watched live streams in Brazil last week, and what was the average viewing duration?”
The AI client calls queryAnalytics with the right metric, dimensions (country = Brazil, content_id), filters (date range), and limit — and returns a structured answer ready for a decision or a report.
No exports. No SQL. No switching tabs. Works the same whether you're using Claude, Cursor, or any other MCP-compatible tool.
Access levels and usage
Read allows viewing your account resources and analytics without any risk of modifying your account — safe to use in shared or exploratory workflows.
Read+Write also allows creating, modifying and deleting resources, across the modules enabled for your account. Start with Read, and grant write access only when a workflow needs it.
When you sign in, the access is what you approve on the consent screen. For tokens, it is the Access level you chose when you created it.
Expiration applies to tokens. After a token expires, MCP requests made with it no longer authenticate. You can choose no expiration or a specific time limit.
Usage tracking Every MCP request counts toward your plan's usage and may be subject to limits depending on your plan. If you reach your limit, requests will be rejected until the quota resets or your plan is upgraded.
Who is this for?
Editorial teams that want to understand content performance without opening analytics every time
Operations teams that monitor live events and need fast answers across streams
Developers building internal tools or automations that combine AI reasoning with Mediastream resources
Product and data teams prototyping AI-assisted workflows on top of real platform data
What's next
The MCP server covers what teams query most: media, live streams, analytics, and ad campaigns, with Read and Read+Write access. Connecting is now a single sign-in with your Mediastream account.
Open platform.mediastre.am/settings/ai, copy the server URL, add it to your AI tool, and test with a simple analytics query. If you don't see the MCP Connection section, contact support via the chat button in Mediastream Platform.