Kinescope MCP server

Connect Claude, Cursor or Qwen3 to your video library. The agent finds videos, edits descriptions and chapters, pulls view reports and schedules live streams from a plain-language prompt.

An MCP server makes video work conversational

You describe the task in chat, and the agent picks the right Kinescope tools and does it: finds videos, edits descriptions, pulls statistics, creates live events. It talks to Kinescope over MCP (Model Context Protocol), an open protocol for connecting AI assistants to services.

AI client

  • Claude Desktop
  • Claude Code
  • Cursor
  • Qwen3
  • Any MCP client

Kinescope MCP

api.kinescope.io/mcp

Streamable HTTP

API token access

Kinescope

  • Videos and metadata
  • Projects, folders, playlists
  • View analytics
  • Live streams and restreaming
  • Subtitles and chapters

What the agent does in Kinescope

These prompts come from the docs. Click one to copy it into your chat.

  • Videos

    Search by title, description and tags, edit metadata and privacy, copy and move videos between projects.

  • Projects and playlists

    The agent creates projects, sorts videos into folders and builds playlists.

  • Analytics

    Views, unique viewers, watch time, completion and retention, geography, devices and drop-off points.

  • Live streams

    Events, schedules, restreaming and recordings.

  • Subtitles and chapters

    The agent adds subtitle files, reads the subtitle text with timestamps and sets chapters.

  • Everything else

    Posters, supplementary materials, Speak rooms, usage and billing statistics.

Uploading video files through MCP isn’t available yet. Use the API for uploads.

Bulk edits in one prompt

The agent finds the videos that match your condition and changes each one. Pick an operation to see it run.

Prompts are taken from the docs. The agent can’t delete videos.

Kinescope plus your other tools

Connect another MCP server to the same agent, for example one for Google apps. The agent takes videos and statistics from Kinescope and moves them into a sheet or a doc, or checks them against your calendar.

These scenarios need a second MCP server for Google apps. Kinescope doesn’t provide it; the data in the window is an example.

Connect in three steps

The server runs at https://api.kinescope.io/mcp. You need a Kinescope account and a desktop AI client.

  1. Step 1

    Create an API token

    In Kinescope, open API tokens, click “Create API token”, name it and choose its permissions. Read-only access is enough for reports.

    Create a token
  2. Step 2

    Add the server to your client

    Paste the config into your client and replace YOUR_API_TOKEN with the token from step 1. Cursor installs it with one button, Claude Desktop and Claude Code with a terminal command.

  3. Step 3

    Check the connection

    Restart the client and send this prompt. The agent replies with a list of your projects.

    Show my projects in Kinescope

    If it doesn’t connect
Add to Cursor

Cursor opens the install dialog for the Kinescope server. Then replace YOUR_API_TOKEN with your token in the MCP settings. Or paste the config by hand:

File ~/.cursor/mcp.json on macOS and Linux, %APPDATA%\Cursor\mcp.json on Windows.

mcp.json
{
  "mcpServers": {
    "Kinescope MCP": {
      "type": "streamable-http",
      "url": "https://api.kinescope.io/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_TOKEN"
      }
    }
  }
}

Ready-made workflows

Send the prompts one by one in the same chat: the agent remembers the result of the previous step.

  • Weekly report

    A summary of the week: top videos, weak spots, new uploads.

    1. Show view statistics for all videos over the last 7 days
    2. Which 5 videos got the most views this week?
    3. Which videos have a completion rate below 60%?
    4. How many videos were uploaded this week and what are their initial metrics?
  • Content audit

    Find the videos that need an update.

    1. Show all videos without a description
    2. Find videos that have no tags
    3. Show videos longer than 10 minutes that have no chapters
    4. Find videos uploaded more than a year ago that haven’t been viewed in the last 3 months
  • Library cleanup

    Sort out projects, tags and privacy.

    1. Show all projects and the number of videos in each
    2. Find videos with identical titles
    3. Show all public videos and determine which ones should be private
    4. Suggest a project and folder structure based on current tags and content

All prompt examples in the docs

Security and limits

No deletion

Delete operations are switched off in the MCP server. The agent can’t delete a video, project or stream, even if asked to.

Token permissions

The agent can do only what your API token allows. Give a read-only token for reports and audits.

Logged like the web app

Every action through MCP lands in your Kinescope account logs: what was viewed or changed, when and from which IP.

Revoke in one step

Remove or disable the server in your AI client, and MCP access stops at once.

Free on any plan

The MCP server costs nothing extra. Your plan’s limits and API rate limits apply.

Public beta

Core features are stable, the toolset is still growing. Direct video upload through MCP is in development.

FAQ

Can’t find the answer you need? Check out our MCP docs.

The Kinescope MCP server is a Model Context Protocol server at https://api.kinescope.io/mcp. Through it, an AI assistant finds and edits videos, manages projects and playlists, pulls analytics and runs live streams in your Kinescope account. It is in public beta.

MCP (Model Context Protocol) is an open protocol, published by Anthropic, for connecting AI assistants to external services. A service describes its actions as tools, and the assistant calls them while it works on a task from the chat.

Cursor, Claude Desktop, Claude Code and the Qwen3 desktop app work with the Kinescope MCP server. Any MCP client that lets you send a header with an API token will do. Web versions such as claude.ai and qwen.chat don’t work, because they can’t store a token.

The Kinescope MCP server is free and works on any Kinescope plan, including the free one. Your plan’s limits apply. Paid plans start at €10/month, see pricing. If your AI client is paid, its subscription is separate; the Qwen3 desktop app is free.

Connect your agent to your video library

Create an account, get an API token and add the server to Cursor, Claude or Qwen3. No credit card required.