Vision

MCP Server

The Reka Vision MCP server lets agent clients such as Claude Code work with your video library. Agents can upload videos, index them for search and Q&A, find relevant moments, inspect transcripts and captions, and ask focused questions about video content.

Use MCP when you want an agent to research across videos from a local workspace or chat client, without wiring each Vision API endpoint into the agent yourself.

Prerequisites

Before you start, make sure you have:

  • A Reka API key from platform.reka.ai
  • Python 3.12 or later
  • uv installed so you can run the server with uvx

Check your local versions:

$python --version
$uv --version

Configure your MCP client

MCP clients start the Reka MCP server from their configuration. Add the server to the client you want to use.

Claude Code

Register the server from the command line:

$claude mcp add reka-mcp -e REKA_VISION_API_KEY=YOUR_API_KEY -- uvx reka-mcp

Claude Desktop

Add the server to ~/Library/Application Support/Claude/claude_desktop_config.json:

1{
2  "mcpServers": {
3    "reka-mcp": {
4      "command": "uvx",
5      "args": ["reka-mcp"],
6      "env": {
7        "REKA_VISION_API_KEY": "YOUR_API_KEY"
8      }
9    }
10  }
11}

Restart Claude Desktop after updating the config.

Codex

Register the server from the command line:

$codex mcp add --env REKA_VISION_API_KEY=YOUR_API_KEY reka-mcp -- uvx reka-mcp

Cursor

Add the server to .cursor/mcp.json:

1{
2  "mcpServers": {
3    "reka-mcp": {
4      "command": "uvx",
5      "args": ["reka-mcp"],
6      "env": {
7        "REKA_VISION_API_KEY": "YOUR_API_KEY"
8      }
9    }
10  }
11}

Verify installation

After configuring your MCP client, restart it and confirm that the Reka MCP server is available.

ClientHow to verify
Claude CodeRun claude mcp list and confirm reka-mcp is listed.
CodexRun codex mcp list, restart your Codex session, and confirm the Reka MCP tools are available.
Claude DesktopOpen the tools menu from the chat input and confirm the Reka tools are available.
CursorOpen MCP settings and confirm reka-mcp is enabled without errors.

Updating

Check which version you’re running:

$uvx reka-mcp --version

To update to the latest version, clear the cached package and restart your MCP client:

$uv cache clean reka-mcp

Available tools

With the MCP server connected, your agent can:

  • Manage videos: Upload, list, inspect, update, and delete videos
  • Organize video libraries: Create, list, and delete video groups
  • Prepare videos for analysis: Index videos for search, Q&A, or full analysis
  • Research and search: Find timestamped moments across indexed videos
  • Understand video content: Ask visual questions about one or more videos
  • Inspect source context: Read transcripts, captions, scenes, and feature status
  • Detect objects: Locate prompted visual elements in short video segments
ToolUse it to
upload_videoUpload a video from a URL or local file path
list_videosList videos in your account or in a video group
get_videoCheck video metadata, upload status, and feature status
update_videoUpdate video name, title, description, or group
delete_videoPermanently delete a video and its indexed data
create_groupCreate a video group
list_groupsList video groups
delete_groupDelete a video group without deleting its videos
index_videoPrepare a video for search, Q&A, or full analysis
search_videosFind timestamped moments across indexed videos
ask_videoAsk questions about one or more videos with visual analysis
get_transcriptRead spoken words as text, segments, or words
get_captionsRead AI-generated visual descriptions with timestamps
get_scenesGet detected scene boundaries
segment_videoDetect objects in short video segments with bounding boxes
get_feature_catalogList available video analysis features and dependencies
summarize_videoGet a compact overview of video status and available context

The MCP server is open source at github.com/reka-ai/reka-mcp.