Dynamic MCP Loader

Load MCP tools at runtime from HTTP/SSE services or local processes.

6|2|Updated Jan 23, 2026
One-click install
npx skills add https://github.com/valkryhx/google_adk_agent --skill dynamic-mcp-loader
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: Dynamic MCP Loader
Source: https://github.com/valkryhx/google_adk_agent/tree/main/skills/dynamic-mcp
Command: npx skills add https://github.com/valkryhx/google_adk_agent --skill dynamic-mcp-loader

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill empowers Agents with the ability to dynamically load new MCP (Model Context Protocol) tools at runtime, eliminating the need for service restarts or configuration changes. Agents can discover and integrate new functionalities on demand, enabling them to handle tasks involving unfamiliar services.

Core Features & Use Cases

  • Runtime Tool Loading: Load MCP tools from remote HTTP/SSE services or by launching local processes.
  • Just-in-Time Tooling: Agents can "download" skills as needed, reducing the overhead of maintaining a large, static toolset.
  • Use Case: When a user requests an action involving an unknown MCP service (e.g., "Use context7 to look up documentation"), the Agent can search for the service's connection details and use connect_mcp to load it, immediately gaining the ability to fulfill the request.

Quick Start

Use the connect_mcp tool to load a remote MCP service from the URL https://mcp.example.com/mcp.

Frequently Asked Questions about Dynamic MCP Loader

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I load MCP tools dynamically at runtime without restarting the agent?

You can load MCP tools dynamically at runtime by using a tool like connect_mcp to link to remote HTTP/SSE services or launch local processes, extending agent capabilities on demand without restarts.

What is just-in-time tooling for agent capabilities and when is it needed?

Just-in-time tooling allows agents to discover and integrate unfamiliar MCP services on demand. It is needed when handling tasks involving unknown services discovered via web search, reducing static toolset overhead.

Can I connect an agent to a remote HTTP/SSE service to expand its skills on demand?

Yes, agents can connect to remote HTTP/SSE services to expand their skills on demand. By fetching connection details and invoking the connect_mcp function, the agent immediately gains the requested functionalities.

How do I integrate a newly discovered MCP service into my running agent?

To integrate a newly discovered MCP service, search for the service's connection details and use connect_mcp with the target URL, allowing the running agent to load the new functionalities immediately.

Does dynamic MCP loading require configuration changes to handle unknown services?

No, dynamic MCP loading eliminates the need for configuration changes or service restarts. Agents can handle unknown services by discovering and integrating new functionalities directly at runtime.