rossai-apify-mcp
MCP server for the public Apify actors published by appreciative_duct. Each tool starts an Actor run with the Apify API and returns the dataset items.
The server reads APIFY_TOKEN from the environment. That token has to be yours. Nothing in this package embeds one.
Tools
A tool waits up to waitSeconds (default 180, maximum 300) and returns at most 100 dataset items. The response includes the run id and the Console URL when there is more to open.
stdio
npm install
npm run build
APIFY_TOKEN=your_token node dist/index.js
Claude Desktop and other stdio clients can use:
{
"mcpServers": {
"rossai-apify": {
"command": "node",
"args": ["/absolute/path/to/rossai-apify-mcp/dist/index.js"],
"env": { "APIFY_TOKEN": "your_token" }
}
}
}
Streamable HTTP
npm run build
APIFY_TOKEN=your_token node dist/index.js --http
The server listens on http://127.0.0.1:8787/mcp. Set PORT to change it. GET /health returns { "ok": true }.
Zero-install Apify MCP
Apify hosts an MCP endpoint that can call these actors without this package. Your Apify token is still required by that host:
Registry files
server.json names the server io.github.rossbuilds/rossai-apify-mcp in the official MCP registry and points at the .mcpb bundle attached to the GitHub release v1.0.0. glama.json lists the maintainer. mcpb/manifest.json is the bundle manifest. The bundle asks for your token at install time. youtube-transcript-batch is not a tool here because that actor stayed private.
Source
https://github.com/rossbuilds/rossai-apify-mcp
Docker
docker build -t rossai-apify-mcp .
docker run -i --rm -e APIFY_TOKEN rossai-apify-mcp
The image runs the stdio server as the non-root node user. No token is baked in; pass APIFY_TOKEN at runtime.