home / mcp / mcp omni connect mcp server
Provides a backward-compatible MCP client with HTTP and stdio transports to connect to MCP servers and manage tools and memory.
Configuration
View docs{
"mcpServers": {
"abiorh001-mcp_omni_connect": {
"url": "http://localhost:8080/mcp"
}
}
}The MCP Omni Core Client provides backward-compatible access to Model Context Protocol (MCP) servers via HTTP and stdio transports. It lets you connect to external MCP tools and services, manage tool access across multiple servers, and orchestrate interactions from a single client. This guide covers practical usage, installation steps, and essential configurations so you can start connecting to MCP servers and using MCP tools right away.
You use the MCP client to connect to MCP servers and access tools exposed by those servers. Start by selecting the transport that matches your MCP server: a standard HTTP-based MCP server or a local (stdio) MCP server running in-process. After connecting, you can list available tools, prompt the MCP client to run prompts or queries, and stream or inspect results. This workflow enables you to access filesystem tools, GitHub tools, or any MCP endpoint, all from one client.
# Option 1: install via uv (recommended)
uv add omnicoreagent
# Option 2: install via pip
pip install omnicoreagentMCP server connections you can use with the client include both remote HTTP endpoints and local stdio servers. The client supports multiple transport types and can switch between them as needed. When you configure a server, provide the appropriate transport, endpoint URL for HTTP servers, or the local command for stdio servers.
Tools available through MCP servers can be discovered and used alongside locally registered tools. You can inspect the full tool list, prompt management, and resource access, making it easy to integrate external capabilities into your automation workflows.
Security and authentication options include bearer tokens and OAuth flows for HTTP MCP servers. For OAuth, an authentication callback server may start automatically during setup. If you don’t need OAuth, you can rely on Bearer tokens via request headers.
Operational tips: start with a simple local stdio server to experiment, then add a remote HTTP MCP server for production workflows. You can register multiple MCP servers and combine them to extend your agent’s capabilities.
The following tools are described as examples you can implement or register in your MCP client workflow.
- get_weather: Retrieve weather information for a given city.
- analyze_data: Analyze a data string and return insights.
- monitor_system: Report current system resource usage.
Get weather information for a specified city.
Analyze provided data and return insights.
Monitor and report current system resources.