Start with AI
AI coding agents can build VideoSDK applications more reliably when they have two things: instructions for how to build with VideoSDK and access to the current VideoSDK APIs while they work.
VideoSDK provides both. The Agent Skill guides the implementation - from understanding the use case to handling authentication and choosing the right features - while the MCP Server lets the agent search the latest VideoSDK documentation whenever it needs an exact API or SDK detail.
The quickest way to set everything up is with the VideoSDK CLI.
Quick Start
1. Install the VideoSDK CLI
- npm
- curl
npm install -g @videosdk.live/cli
curl -fsSL https://videosdk.live/install-cli | bash
The CLI requires Node.js 20 or later. The curl install works on macOS and Linux only. On Windows, use npm.
See the CLI overview for installation details and available commands.
2. Install the VideoSDK Agent Skill
The VideoSDK Agent Skill gives supported AI coding agents the instructions they need to build VideoSDK applications using recommended workflows and authentication patterns.
- VideoSDK CLI
- npx skills
vsdk skills
Choose the agents where you want to install the skill, then restart the agent.
npx skills add videosdk-live/skills
npx skills works across supported coding agents. Select the agent you use, such as Claude Code, Cursor, or Codex, then restart it after installation.
3. Connect the MCP Server
The server is at https://docs.videosdk.live/mcp and speaks streamable HTTP. Pick your client:
- VideoSDK CLI
- Claude Code
- Antigravity CLI
- Codex CLI
- Cursor
- VS Code
Name your client and the CLI runs the add command or writes the config file for you:
vsdk mcp --client claude-code --install
--client takes claude-code, cursor, vscode, windsurf, antigravity, codex, zed, or cline.
Run this from your project, or add --scope user to make the server available everywhere:
claude mcp add --transport http videosdk https://docs.videosdk.live/mcp
Confirm it registered with claude mcp list.
Add the server to ~/.gemini/config/mcp_config.json for every project, or to .agents/mcp_config.json for this one:
{
"mcpServers": {
"videosdk": {
"serverUrl": "https://docs.videosdk.live/mcp"
}
}
}
A remote server must use serverUrl. The legacy url and httpUrl fields are not supported.
Add the server to ~/.codex/config.toml:
[mcp_servers.videosdk]
url = "https://docs.videosdk.live/mcp"
Add the server to ~/.cursor/mcp.json for every project, or to .cursor/mcp.json for this one:
{
"mcpServers": {
"videosdk": {
"url": "https://docs.videosdk.live/mcp"
}
}
}
Create .vscode/mcp.json in your workspace:
{
"servers": {
"videosdk": {
"type": "http",
"url": "https://docs.videosdk.live/mcp"
}
}
}
Copilot picks the server up once the file is saved.
Most clients need a restart before a newly added server initialises. The MCP Server page covers verification and troubleshooting.
4. Start Building
Tell the agent which platform you are using and what you want to build. Include any important roles, features, or application requirements.
If something important is missing, the agent can ask for it before starting the implementation.
Use the VideoSDK skill to add video calling to this Next.js app, with a server route that mints the auth token.
Build a telehealth consultation in Flutter, doctor and patient, with a waiting room.
Build a live shopping stream in React where viewers can be brought up on stage.
Verify Your Setup
Run a VideoSDK task and confirm that the agent:
- asks for important missing requirements;
- uses
docs.videosdk.livefor VideoSDK API details; - keeps your API secret out of client-side code.
If these work, your setup is ready.
Agent Skill vs. MCP Server
The Agent Skill and MCP Server solve different parts of the development workflow and work best together.
| Agent Skill | MCP Server | |
|---|---|---|
| Purpose | Guides the agent through building with VideoSDK | Gives the agent access to VideoSDK documentation |
| Provides | Scoping, authentication patterns, implementation guidance, and feature selection | Current SDK documentation searchable by SDK and version |
| Used when | Planning and implementing the application | Looking up APIs, methods, configuration, and SDK details |
The Agent Skill guides how the application should be built. When the agent needs an exact VideoSDK API or implementation detail, it can retrieve it through the MCP Server.
Every SDK also includes an AI Tools section with platform-specific resources and prompts.
If Your Tool Does Not Support MCP
You can still give an AI coding agent access to VideoSDK documentation without connecting an MCP server.
Use the VideoSDK CLI
If your agent can run shell commands, it can search and read VideoSDK documentation directly from the terminal.
Search the documentation:
vsdk docs search "<question>" --sdk <sdk>
Then retrieve a page as Markdown:
vsdk docs get <url>
The CLI searches the same VideoSDK documentation used by the MCP Server.
See VideoSDK CLI documentation commands for more details.
Use the LLM Documentation Files
For tools that can read URLs or text files directly, VideoSDK provides AI-friendly documentation indexes.
https://docs.videosdk.live/llms.txt
Lists VideoSDK documentation pages along with their descriptions, making it useful for discovering the right page.
https://docs.videosdk.live/llms-full.txt
Contains the complete VideoSDK documentation in a single file for tools that can work with larger context files.
Every documentation page also includes a copy option in the top-right corner, allowing you to copy the cleaned Markdown or open it with supported AI tools such as ChatGPT or Perplexity.
