WAVE MCP Server
wave-av/mcp-server
APIs de streaming WAVE como tools MCP para Claude, Cursor e Windsurf
Ver no GitHubServidor MCP que expõe APIs de streaming WAVE como ferramentas para assistentes de IA (Claude, Cursor, Windsurf). Fornece 19 tools para gerenciar streams, produções em estúdio, analytics e billing com suporte a recursos via URI scheme wave://.
Instalação
npx convez-ai wave-av/mcp-server1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Gerenciar streams e produções multi-câmera via IA
- Monitorar métricas de saúde e performance de streams em tempo real
- Automatizar criação e controle de streams em fluxos de IA
- Acessar dados de viewers e usage de forma programática
Veredito

MCP server bem estruturado que integra plataforma WAVE de streaming com modelos de IA, oferecendo cobertura funcional sólida (streams, produções, analytics, billing). Implementação TypeScript com suporte eficiente em-process e stdio, ideal para automação de workflows de stream/produção.
README
A tradução para o português deste README ainda está na fila.
@wave-av/mcp-server
MCP (Model Context Protocol) server that exposes WAVE streaming APIs as tools for AI coding assistants.
Quick start
npx @wave-av/mcp-server
Setup
1. Get an API key
# Via CLI
wave auth login
# Or create at https://wave.online/settings/api-keys
2. Configure your AI tool
Add to your .mcp.json (Claude Code, Cursor, Windsurf, etc.):
{
"mcpServers": {
"wave": {
"command": "npx",
"args": ["-y", "@wave-av/mcp-server"],
"env": {
"WAVE_API_KEY": "wave_live_..."
}
}
}
}
Available tools
Streams
| Tool | Description |
|---|---|
wave_list_streams | List all streams with pagination and status filtering |
wave_create_stream | Create a new stream with protocol and privacy options |
wave_start_stream | Start streaming on an existing stream |
wave_stop_stream | Stop an active stream |
wave_get_stream_health | Get real-time health metrics for a stream |
Studio
| Tool | Description |
|---|---|
wave_list_productions | List studio production sessions |
wave_create_production | Create a new multi-camera production |
Analytics
| Tool | Description |
|---|---|
wave_get_viewers | Get current viewer count and breakdown |
wave_get_stream_metrics | Get detailed stream performance metrics |
Billing
| Tool | Description |
|---|---|
wave_get_subscription | Get current subscription plan and status |
wave_get_usage | Get current period usage and limits |
Resources
Access WAVE entities directly via the wave:// URI scheme:
wave://streams/{id}- Stream configuration and statuswave://productions/{id}- Studio production details
Environment variables
| Variable | Required | Default | Description |
|---|---|---|---|
WAVE_API_KEY | Yes | - | Your WAVE API key |
WAVE_BASE_URL | No | https://wave.online | API base URL |
In-process (Claude Agent SDK) mode
For consumers already running inside a Claude Agent SDK
session, the same tools are available in-process — skipping the stdio subprocess
hop (~50 ms vs ~500 ms cold start). The tool list is shared with the stdio
server (src/tools/index.ts), so the two transports never drift.
@anthropic-ai/claude-agent-sdk is an optional peer dependency: stdio users
never need it. Install it only for this mode:
npm install @wave-av/mcp-server @anthropic-ai/claude-agent-sdk
import { query } from "@anthropic-ai/claude-agent-sdk";
import { createWaveSdkMcpServer } from "@wave-av/mcp-server/sdk-server";
const wave = await createWaveSdkMcpServer();
for await (const message of query({
prompt: "List my active streams",
options: { mcpServers: { wave }, env: { WAVE_API_KEY: process.env.WAVE_API_KEY } },
})) {
// handle messages
}
Setup for other AI tools
Cursor
Add to .cursor/mcp.json:
{
"mcpServers": {
"wave": {
"command": "npx",
"args": ["-y", "@wave-av/mcp-server"],
"env": { "WAVE_API_KEY": "wave_live_..." }
}
}
}
Windsurf
Add to Windsurf MCP settings with the same configuration.
Troubleshooting
Server not starting
Verify your API key is set:
echo $WAVE_API_KEY
Tools not appearing
Restart your AI tool after adding the MCP configuration. Most tools require a restart to detect new MCP servers.
Connection errors
The MCP server uses stdio transport (no network listener). If you see connection errors, check that npx can run successfully:
npx @wave-av/mcp-server --version
Testing the server
Send a JSON-RPC initialize request to verify:
echo '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0.0"}}}' | npx @wave-av/mcp-server
Related packages
- @wave-av/sdk — TypeScript SDK (34 API modules)
- @wave-av/adk — Agent Developer Kit
- @wave-av/cli — Command-line interface
- @wave-av/create-app — Scaffold a new project
- OpenAPI spec — Full API specification
Development
cd packages/mcp-server
pnpm install
pnpm run build
pnpm run dev # Watch mode
pnpm run type-check
License
MIT