Give your AI agent eyes

LLMs cannot watch video. They cannot process frames, listen to audio, or read on-screen text from a video file. VidContext converts any video into structured text your agent can reason about — via a single API call or the built-in MCP server.

The gap in every AI agent

Your agent can read documents, search the web, query databases, and write code. But hand it a video file and it hits a wall. Video is the largest source of unstructured information on the internet, and your agent is blind to it.

VidContext bridges this gap. It watches the video — every frame, every word spoken, every piece of text on screen — and returns a structured text document your agent can process like any other input. Scenes, transcripts, detected brands, audio analysis, and scoring frameworks. All from one call.

Built for agent workflows

Two ways to integrate

REST API

curl -X POST https://api.vidcontext.com/v1/analyze \
  -H "X-API-Key: vc_your_key_here" \
  -F "file=@video.mp4" \
  -F "output_format=context"

# Returns structured JSON in ~50 seconds
# Feed the response directly into your agent's context

MCP server (for Claude, Cursor, and MCP-compatible agents)

# Install the MCP server
pip install vidcontext-mcp

# Add to your agent's MCP config:
{
  "mcpServers": {
    "vidcontext": {
      "command": "vidcontext-mcp",
      "args": ["--api-key", "vc_your_key_here"]
    }
  }
}

# Your agent can now call VidContext as a tool:
# "Analyze this video and tell me what brands appear"
# The agent handles the rest automatically.

The MCP server exposes VidContext as a tool your agent can call directly. No wrapper code needed.

What agents build with VidContext

Frequently asked questions