Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@zrok-mcp Share my local server on port 3000 publicly via zrok"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
An MCP server for zrok — the open-source secure sharing platform built on OpenZiti. Lets AI agents manage zrok tunnels, shares, and access programmatically.
Supports three transports: stdio (local), streamable-http and SSE (remote/Bifrost).
Each tool uses an action parameter to select the operation — no need to remember many tool names.
zrok_env — Environment managementAction | Description |
| Check environment status (enabled, API endpoint, identity) |
| Enable the zrok environment with an account token |
| Disable the current zrok environment |
zrok_share — Share managementAction | Description |
| Create a public or private share |
| Delete a share by token |
| List all shares with optional filters |
zrok_access — Access managementAction | Description |
| Create access to a private share |
| Remove access to a share |
| List all accesses with optional filters |
Related MCP server: Anycast MCP Server
Python 3.10+
zrok2 Python SDK
A zrok account and enabled environment
cd zrok-mcp
pip install -e .Transport is selected via the ZROK_MCP_TRANSPORT environment variable (default: stdio).
Transport | Use case |
| Local clients (Claude Desktop, Cursor, Crush) |
| Remote / Bifrost gateway (recommended) |
| Remote / Bifrost gateway (legacy SSE) |
zrok-mcpZROK_MCP_TRANSPORT=streamable-http ZROK_MCP_PORT=8000 zrok-mcpdocker build -t zrok-mcp .
docker run -p 8000:8000 -v ~/.zrok2:/root/.zrok2 zrok-mcpmcp dev src/zrok_mcp/server.pyVariable | Default | Description |
|
| Transport: |
|
| Bind host (HTTP/SSE transports) |
|
| Bind port (HTTP/SSE transports) |
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"zrok": {
"command": "uv",
"args": ["--directory", "/path/to/zrok-mcp", "run", "zrok-mcp"]
}
}
}Add to your crush settings or MCP config:
{
"zrok": {
"command": "uv",
"args": ["--directory", "/path/to/zrok-mcp", "run", "zrok-mcp"]
}
}Deploy zrok-mcp as a remote HTTP server (Docker, Railway, etc.)
In the Bifrost Web UI, go to MCP Gateway → New MCP Server
Select HTTP as the connection type
Enter the URL: http://your-zrok-mcp-host:8000/mcp
Set tools_to_execute to ["*"] (or filter as needed)
Or via config file:
{
"mcp": {
"mcp_clients": [
{
"name": "zrok",
"connection_type": "http",
"connection_string": "http://your-zrok-mcp-host:8000/mcp",
"tools_to_execute": ["*"]
}
]
}
}For SSE transport, use connection_type: "sse" and connection_string: "http://your-zrok-mcp-host:8000/sse".
Once configured, your AI agent can do things like:
"Share my local server on port 3000 publicly via zrok"
The agent will call zrok_share(action="create", target="http://localhost:3000") and return the public URL.
"List all my zrok shares"
The agent will call zrok_share(action="list") and show token, mode, target, and endpoints.
"Create a private TCP tunnel for my database on port 5432"
The agent will call zrok_share(action="create", share_mode="private", backend_mode="tcpTunnel", target="localhost:5432").
MIT
This server cannot be deployed
Provides capabilities that let LLM agents perform a range of infrastructure management tasks.
Manage portable AI agent playbooks, Agent Skills, MCP configurations, personas, and memory.
Manage SRG+ hubs, channels, content, assets, users, and workspaces from any MCP-aware AI agent.
Give your AI agents the tools to build, manage, and run automation workflows.
