MCP Server
Let AI assistants analyze videos directly through Claude, Cursor, or Windsurf.
The NetraFlow MCP server lets AI assistants submit and retrieve video analyses directly. It is hosted — there is nothing to install, and no API key to paste. You sign in with your NetraFlow account the first time your client connects, and it works against your workspace.
Connect
Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"netraflow": {
"url": "https://mcp.netraflow.com/mcp"
}
}
}Restart the app, then approve the NetraFlow sign-in prompt.
Add to .cursor/mcp.json in your project root:
{
"mcpServers": {
"netraflow": {
"url": "https://mcp.netraflow.com/mcp"
}
}
}Add to ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"netraflow": {
"url": "https://mcp.netraflow.com/mcp"
}
}
}Clients that only speak the older SSE transport should point at https://mcp.netraflow.com/sse instead.
Available tools
| Prop | Type | Description |
|---|---|---|
| analyze_video | submit | Submit a video URL for analysis. Returns a job_id for tracking. |
| get_analysis | read | Get full analysis results for a completed job. |
| check_status | read | Check processing status and progress of a job. |
| list_analyses | read | List recent video analysis jobs with optional filtering. |
| cancel_job | write | Cancel a pending or processing job. |
| create_batch | submit | Submit a batch of video URLs sharing the same capabilities. |
| get_batch | read | Get batch status, per-job progress, and aggregate results. |
| list_batches | read | List batches with optional filtering. |
| cancel_batch | write | Cancel a batch and its pending or processing jobs. |
| check_credits | read | Check account information and remaining credit balance. |
| get_usage | read | Get usage statistics and transaction history. |
Tools act on the workspace you signed in with, and are subject to the same plan limits as the API — batch tools require a plan with batch enabled.
Following creators
The tools above act on a video that already exists. Follows are the standing version: point one at a YouTube channel or TikTok profile, describe what you care about in plain language, and every new upload is checked against that description. Checks are free — only matches are analyzed and charged.
MCP has no way to push to your assistant, so a match that lands next week cannot reach the conversation that created
the follow. Set notify when you create one and results are delivered to Slack, a webhook, or a teammate's email
instead. Without it, matches surface only in the dashboard and the daily digest.
| Prop | Type | Description |
|---|---|---|
| follow_creator | write | Follow a channel or profile and auto-analyze new videos matching a rule described in plain language. |
| list_follows | read | List follows with their rules, delivery channels, and check/match/analyze counts. |
| update_follow | write | Pause or resume a follow, rewrite its rule, or change delivery. Rule and notify are replaced, not merged. |
| unfollow_creator | write | Stop following a creator. Analyses already produced are kept. |
| get_follow_activity | read | Recent videos a follow checked, with the verdict and reason for each — including the ones it rejected. |
Follow counts and how many videos they may analyze per week are both plan limits — list_follows returns the ceilings that apply to your workspace.