Skip to content
Conduit

mcp server

vmodal-mcp

MCP server for AI Agent for Multimodal Agent

v-modal/vmodal_mcpcommunitylistedunreviewed

Set up

Runs on your computer. Pick your agent and operating system.

Before you start

  • uv (`curl -LsSf https://astral.sh/uv/install.sh | sh`)
~/Library/Application Support/Claude/claude_desktop_config.json
{
  "mcpServers": {
    "vmodal-mcp": {
      "command": "uvx",
      "args": [
        "vmodal_mcp"
      ]
    }
  }
}

same format Cursor, GitHub Copilot CLI, Junie, JetBrains AI Assistant, Augment (Auggie CLI), Factory Droid, Kiro, Devin CLI +12

Values in angle brackets are placeholders. Restart Claude Desktop after editing its config.Claude Desktop docs

README

synced Oct 21, 2026

After setup, just ask:

Find a red car in my city collection.

A collection is a group of your videos or images. A sub-collection is a smaller group inside it. You can search a whole collection without choosing a sub-collection. VModal chooses the search index version automatically.

Set up once

You need Python 3.10 or newer and an AI assistant from the list above. Run the setup commands in your terminal. After setup, type the prompts in your assistant's chat.

1. Get your API key

Sign up for a VModal API key. Copy your key. Wherever you see PUT_API_KEY_HERE below, replace it with your key. Keep your key private.

2. Install VModal

For Claude Code or Codex, install and connect in one command. Choose one and enter your API key when prompted:

curl -fsSL https://raw.githubusercontent.com/v-modal/vmodal_mcp/refs/heads/main/install_mcp.sh | bash -s -- claude
curl -fsSL https://raw.githubusercontent.com/v-modal/vmodal_mcp/refs/heads/main/install_mcp.sh | bash -s -- codex

Requires curl, git, Python with pip, and the selected assistant CLI. You can also export api_key before running the command. After it succeeds, skip to step 4.

For manual installation or another assistant:

Copy this command into your terminal:

python -m pip install --upgrade "git+https://github.com/v-modal/vmodal_mcp.git@main"

3. Connect your assistant

Follow only the instructions for your assistant.

Claude Code

Paste this into your terminal after replacing PUT_API_KEY_HERE:

claude mcp add --scope user vmodal \
  -e api_key=PUT_API_KEY_HERE -- \
  vmodal-mcp run --transport stdio

Codex

Paste this into your terminal after replacing PUT_API_KEY_HERE:

codex mcp add --env api_key=PUT_API_KEY_HERE vmodal -- \
  vmodal-mcp run --transport stdio

GitHub Copilot CLI

Paste this into your terminal after replacing PUT_API_KEY_HERE:

copilot mcp add vmodal --env api_key=PUT_API_KEY_HERE -- \
  vmodal-mcp run --transport stdio

Cursor

In your project, create a file named .cursor/mcp.json. Paste the following into it, replace PUT_API_KEY_HERE with your key, and save:

{
  "mcpServers": {
    "vmodal": {
      "command": "vmodal-mcp",
      "args": ["run", "--transport", "stdio"],
      "env": {
        "api_key": "PUT_API_KEY_HERE"
      }
    }
  }
}

4. Restart your assistant and check the connection

Open a new chat and paste:

Use VModal to check my connection and show my account.

If your assistant shows your account, you're ready.

Start here: copy a prompt into chat

See your collections

List all my collections.

See your sub-collections

List all my sub-collections in a table with two columns:
Collection | Sub-collection

Find something

Replace city with a collection name from your list:

Find a red car.
Collection: city

This searches the whole collection. To search one sub-collection, add its name:

Find a red car.
Collection: city
Sub-collection: rome

Get the matching pictures

Find a red car in my city collection. Save the best three matching
pictures to my computer and tell me where to find them.

Upload your first video

Replace ./my_video.mp4 with the path to a video on your computer:

Upload ./my_video.mp4 to a collection named my_videos, in a sub-collection
named my_uploads. Make it searchable, wait until it's ready, then find
a person walking outside.

More ready-to-copy prompts →

Try it with a sample video (optional)

Run these commands in your terminal, with your API key:

export api_key=PUT_API_KEY_HERE
vmodal test

This uploads an included sample video, makes it searchable, and runs a search. It usually takes about a minute. Look for ok: true in the result. Each run creates a new test collection.

Need help?

  • VModal doesn't appear in your assistant: restart your assistant and open a new chat.
  • Your API key is rejected, or you see 401: check the key you entered in step 3.
  • vmodal-mcp or vmodal is not found: restart your terminal and retry. If it still fails, repeat step 2 using the Python installation your terminal uses.
  • No matches: ask your assistant to list your collections, check the collection name, and try a simpler description. If you just uploaded a video, ask whether it is ready to search.

For developer details, see the tool contract.

↑↓ navigate openesc close